Archive for the ‘Computing Models’ Category

SDDC, SDN, NFV, SFV, ACI, Service Governor, Super Recursive Algorithms and All That Jazz:
October 27, 2014

“It’s very likely that on the basis of philosophy that every error has to be caught, explained, and corrected, a system of the complexity of the living organism would not run for a millisecond.“
—–   von Neumann, Papers of John von Neumann on Computing and Computing Theory, Hixon Symposium, September 20, 1948, Pasadena, CA, The MIT Press, 1987.

Communication, Collaboration and Commerce at the Speed of Light:

With the advent of many-core servers, high bandwidth network technologies connecting these servers, and new class of high performance storage devices that can be optimized to meet the workload needs (IOPs intensive, throughput sensitive or capacity hungry workloads), Information Technology (IT) industry is looking at a transition from its server-centric, low-bandwidth, client-server origins to geographically distributed, highly scalable and resilient composed service creation, delivery and assurance environments that meet the rapidly changing business priorities, latency constraints, fluctuations in workloads and availability of required resources. Distributed service composition and delivery brings new challenges with scale and fluctuations both in demand and the availability of resources. New approaches are emerging to improve resiliency and the efficiency of distributed system design, deployment, management and control.

The Jazz Metaphor:

The quest for transition is best described by the Jazz metaphor aptly summarized by Holbrook [1] (Holbrook 2003), “Specifically, creativity in all areas seems to follow a sort of dialectic in which some structure (a thesis or configuration) gives way to a departure (an antithesis or deviation) that is followed, in turn, by a reconciliation (a synthesis or integration that becomes the basis for further development of the dialectic). In the case of jazz, the structure would include the melodic contour of a piece, its harmonic pattern, or its meter…. The departure would consist of melodic variations, harmonic substitutions, or rhythmic liberties…. The reconciliation depends on the way that the musical departures or violations of expectations are integrated into an emergent structure that resolves deviation into a new regularity, chaos into a new order, and surprise into a new pattern as the performance progresses.”

The Thesis:

The thesis in the IT evolution is the automation of business processes and service delivery using client-server architectures. It served well as long as the service scale and fluctuations of service delivery infrastructure resources were within certain bounds that allowed the action to increase or decrease available resources and meet the fluctuating demands. In addition, the resiliency of the service is always adjusted by improving the resiliency (availability, performance and security) of the infrastructure through various appliances, processes and tools. This introduced a timescale for meeting the resiliency required for various applications in terms of recovery time objectives and recovery point objectives. The resulting management “time constant” (defined as the time to recover a service to meet customer satisfaction) has been continuously decreasing with the use of newer technologies, tools and process automation.
However, with the introduction of the high-speed Internet, access to mobile technology and globalization of e-commerce, the scale and fluctuations in service demand have radically changed which have put challenging demands on provisioning the resources within shorter and shorter periods of time. Figure 1 summarizes the key drivers that are forcing the drastic reduction of management time constant.

Business Drivers for Anti-Thesis

Figure 1: Global communication, collaboration and commerce at the speed of light is forcing the drastic reduction in IT resource management time constant

 The Anti-Thesis:

The result is the anti-thesis (the word is not used pejoratively but actually it denotes innovation, creativity and a touch of anti-establishment rebellion in the Jazz metaphor) to virtualize the infrastructure management (compute, storage and network resources) and provide intelligent resource management services that utilize commodity infrastructure connecting fat pipes. Software defined data center (SDDC) is used to represent the dynamic provisioning of  server clusters connected by a network attached to required storage all meeting the service levels required by the applications that are composed to create a service transaction. The idea is to monitor the resource utilization by these service components and adjust the resources as required to meet the Quality of Service (QoS) needs of the service transaction (in terms of cpu, memory, network bandwidth, latency, storage throughput, IOPs and capacity.) Network function virtualization (NFV) is used to denote the dynamic provisioning and management of network services such as routing, switching and controlling commodity hardware that is solely devoted to connect various devices to assure desired network bandwidth and latency. Storage function virtualization (SFV) similarly denotes the dynamic provisioning and management of commodity storage hardware with required IOPs, throughput and capacity. ACI denotes application centric infrastructure which is sensitive to the needs of particular application and dynamically adjusts the resources to provide right cpu, memory, bandwidth, latency, storage IOPs, throughput and capacity. The drive to move away from proprietary network and storage equipment to commodity high performance hardware made ubiquitous with open interface architectures are intended to foster competition and innovation both in hardware and software. The open software is supposed to match the needs of the application by tuning the resources dynamically using the compute, network and storage management function made available with open-source software.

Unfortunately, the anti-thesis brings its own issues in transforming the current infrastructure that has evolved over few decades to the new paradigm.

  1. The new approach has to accommodate current infrastructure and applications and allow seamless migration to new paradigm without vendor lock-in to use new infrastructure. Fork-lift strategy will not work that involves time. money and service interruption.
  2. Current infrastructure is designed to provide low latency high performance application quality of service with various levels of security. For mission critical applications to migrate to new paradigm, these requirements have to be met without compromise.
  3. The new paradigm should not require new way of developing applications or it must support current development languages and processes without new methodology lock-in. An application is defined both by functional requirements that dictate the specific domain functions and logic as well as non-functional requirements that define operational constraints related to service availability, reliability, performance, security and cost dictated by business priorities, workload fluctuations and resource latency constraints. A non-functional requirement specifies criteria that can be used to judge the operation of a system, rather than specific behaviors. The plan for implementing functional requirements is detailed in the system design. The plan for implementing non-functional requirements is detailed in the system architecture. The architecture for non-functional requirements plays a key role in whether the open systems approach will succeed or fail. An architecture that defines a plug and play approach requires a composition scheme which leads to the next issue.
  4. There must be a way to compose applications developed by different vendors without having to look inside their implementation. In essence there must be a composition architecture that allows applications to be developed independently but can be composed to create new applications without having to modify the original components. Even when you have open-sourced applications, integrating them and creating new workflows and services is a labor intensive and knowledge sensitive task. The efficiency will be thwarted by the need for service engagements, training and maintenance of integrated workflows.

Current approaches suggested in the anti-thesis movement embracing virtual machines (VM), open-sourced applications and cloud computing fail on all these accounts by increasing complexity or requiring vendor, API and architecture dependency. The result is increased operation cost of integration dependency on ad-hoc software and services.

The increase in complexity with scale and distribution is more an issue of architecture and is not addressed by throwing more ad-hoc software to automate with managers of managers, point solutions and tools. It has to do more with the limitation of current computing architecture than lack of good ad-hoc software approaches.

Server virtualization creates a Virtual Machine image that can be replicated easily in different physical servers with shared resources. The introduction of Hypervisor to virtualize hardware resources (cpu and memory) allows multiple virtual machine images to share the resources in a physical server. NFV and SFV provide management functions to control the underlying commodity hardware. OpenStack and other infrastructure provisioning mechanisms have evolved through the anti-thesis movement to integrate VM provisioning integrated with NFV and SFV provisioning to create clusters of VMs on which the applications can deliver the service transactions. Figure 2 shows OpenStack implementation of such a service provisioning process. A cluster of VMs required for a service delivery can be provisioned with required service level agreements to assure right cpu, memory, bandwidth, latency, storage IOPs, throughput and capacity. It is also important to note that OpenStack not only can provision a VM cluster but also physical server cluster or a mixture. It allows adding or deleting or tuning a VM on demand. In addition, OpenStack allows including applications themselves to be part of the image and snapshots that can be reused to replicate the VM on any server. Clusters with appropriate applications and dependencies with connectivity and firewall rules can be provisioned and replicated. This allows for orchestration of VM images to provide auto-failover, auto-scaling, live-migration and auto-protection for service delivery.

OpenStack based infrastructure control plane

Figure 2: OpenStack is used to provision infrastructure with required service level agreements to assure cpu, memory, bandwidth, storage IOPs, throughput, storage capacity of individual virtual machine (VM) and the network latency of the VM cluster

Unfortunately, the anti-thesis movement solely depends on infrastructure mobility and management through VMs and associated plumbing which requires a lock-in on the availability of same OpenStack in a distributed environment or complex image orchestration add-ons. More recently instead of moving the whole virtual image containing the OS, run-time environments and applications along with their configurations, a mini-OS (using subset of operating system services) image is created with application and their configurations. LXC containers and Docker containers are examples. The use of mobility of VMs or containers to move applications from one infrastructure to another to manage the infrastructure SLAs to meet QoS needs of an application has created a plethora of ad-hoc solutions adding to the complexity. Figure 3 shows the current state-of-the-art.

Anti-Thesis: urrent State of the Art

Figure 3: Current state-of-the-art that provides application QoS through Virtual Machine mobility or container mobility where container is also an image

While this approach provides a solution to meet application scaling and fluctuations needs as long as the infrastructure meets certain requirements, there are certain shortcomings in distributed heterogeneous infrastructures provided by different vendors:

  1. Multiple Orchestrators are required when different architectures and infrastructure management systems are involved
  2. Too many infrastructure management tools, point solutions and integration services increase cost and complexity
  3. Manager of Managers create complexity
  4. Cannot scale across distributed infrastructures belonging to different service providers to leverage co0mmodity infrastructures resulting vendor lock-in
  5. VM Image Mobility creates additional VM image management, run away bandwidth and storage with proliferation of VM instances
  6. Lack of end-to-end service security visibility and control when services span across multiple service infrastructures.
  7. Managing low-latency transactions in distributed environments increases cost and complexity

Figure 4 shows the complexity involved in scaling services across distributed heterogeneous infrastructures with different owners using different infrastructure management systems. Integrating multiple distributed infrastructures with disparate management systems is not a highly scalable solution without increasing complexity and cost.

Obviously if scale, distribution and fluctuations (both in demand and resources) are not a requirement, then, the thesis will do well. Today, there are still many main-frame systems providing high transaction rates albeit at a higher cost. Anti-thesis is born out of the need for high degree of scalability, distribution and fluctuations with higher efficiency. Big data analysis, large scale collaboration systems are examples. However there is a large class of services that like to leverage commodity infrastructure and resiliency with security and application QoS management without vendor lock-in or high cost of complexity.

There are three stakeholders in an enterprise who want different things from infrastructure to provide QoS assurance:

  1. The Line of business owners and the CIO want:
    1. Service Level Quality (availability, performance, security and cost) Assurance
    2. End-to-end service visibility and control
    3. Precise resource accounting
    4. Regulatory Compliance
  2. The IT infrastructure providers want:
    1. Provide “Cloud-like Services” in private datacenters
    2. Advantage of commodity infrastructure without vendor lock-in
    3. Ability to “migrate service” or “tune infrastructure SLAs” based on Policies and application demand
    4. Ability to burst into cloud without vendor-lock-in
  3. The developers want:
    1. Focus on business logic coding and specification of run-time requirements for resources (application intent, context, communications, control and constraints) without worrying about run-time infrastructure configurations
    2. Converged DevOps to develop test and deploy with agility
    3. Service deployment architecture decoupling non-functional and functional requirements
    4. Service composition tools for reuse
    5. End-to-end visibility and profiling at run-time across the stack for Debugging

In essence, service developers would want to focus on functional requirement fulfillment without having to worry about resource availability in a fluctuating environment. Monitoring resource utilization and taking action on non-deterministic impact of scaling and fluctuations should be supported by a common architecture that decouples application execution from underlying resource management distributed or not.

Complexity and Cost

Figure 4: Complexity in a distributed infrastructure where scaling and fluctuations are increasing

The Synthesis:

The synthesis depends on addressing the scaling and fluctuation issues without vendor lock-in or architecture lock-in that restricts developers to use their current environments and requires accommodating current infrastructure while allowing new infrastructure with NFV and SFV to seamlessly integrate. For example the anti-thesis solutions require certain features in their OSs and new middleware must run in distributed environments. This leaves a host of legacy systems out.

A call for the synthesis is emerging from two quarters:

  1. Industry analysts such as Gartner who predict that a service governor will emerge in due time. “A service governor [2] is a runtime execution engine that has several inputs: business priorities, IT service descriptions (and dependency model), service quality and cost policies. In addition, it takes real-time data feeds that assess the performance of user transactions and the end-to-end infrastructure, and uses them to dynamically optimize the consumption of real and virtual IT infrastructure resources to meet the business requirements and service-level agreements (SLAs). It performs optimization through dynamic capacity management (that is, scaling resources up and down) and dynamically tuning the environment for optimum throughput given the demand. The service governor is the culmination of all technologies required to build the real-time infrastructure (RTI), and it’s the runtime execution management tool that pulls everything together.”
  2. From the academic community who recognize the limitations of Turing’s formulation of computation in terms of functions to process information using simple read, compute (change state) and write instructions combined with the introduction of program, data duality by von Neumann which has allowed information technology (IT) to model, monitor, reason and control any physical system. Prof. Mark Burgin [3] in his 2005 book on super recursive algorithms states “it is important to see how different is functioning of a real computer or network from what any mathematical model in general and a Turing machine,(as an abstract, logical device), in particular, reputedly does when it follows instructions. In comparison with instructions of a Turing machine, programming languages provide a diversity of operations for a programmer. Operations involve various devices of computer and demand their interaction. In addition, there are several types of data. As a result, computer programs have to give more instructions to computer and specify more details than instructions of a Turing machine. The same is true for other models of computation. For example, when a finite automaton represents a computer program, only certain aspects of the program are reflected. That is why computer programs give more specified description of computer functioning, and this description is adapted to the needs of the computer. Consequently, programs demand a specific theory of programs, which is different from the theory of algorithms and automata.”

In short, the programs (or functions) developers develop to code business logic do not contain knowledge about how compute, storage and network devices interact with each other (structure) and how to deal with changing business priorities, workload variations and latency constraints (fluctuations that force changes to structure). This knowledge has to be incorporated in the architecture of the new computing, management and programming model.

These non-functional requirements are requirements that specify criteria that can be used to judge the operation of a system, rather than specific behavior. This should be contrasted with functional requirements that define specific behavior or functions that deal with algorithms, or business logic. The plan for implementing functional requirements is detailed in the system design. The plan for implementing non-functional requirements is detailed in the system architecture. These requirements include availability, reliability, performance, security, scalability and efficiency at run-time. The new architecture must encapsulate the intent of the program, its operational requirements such as the context, connectivity to other components, constraints and control abstractions that are required to manage the non-functional requirements. Figure 5 shows an architecture where the service management architecture is decoupled from the infrastructure management systems monitoring and managing distributed resources that may belong to different providers with different incentives.

Infusing Cognition into Service Control Plane

Figure 5: A cognition infused service composition architecture that decouples distributed heterogeneous multi-vendor infrastructure management

The infrastructure control plane provides automation, monitoring and management of infrastructure required for applications to execute their intent. The output of the infrastructure is a cluster of physical servers or virtual servers with an operating system in each server to provide well-defined computing resources in terms of total CPU, Memory, network bandwidth, latency, storage IOPs, throughput and capacity. The infrastructure control plane will be able to provide required clusters on demand and elastically scale the nodes or the individual node resources on demand. The elastic on-demand resources use automation processes or NFV and SFV resources connected to Virtual or Physical servers.

As Professor Mark Burgin points out, the intent and the application monitoring to process information, apply knowledge, and change the circumstance must be part of the service management knowledge independent of distributed infrastructure management systems for providing true scalability, distribution and resiliency; and avoiding vendor lock-in or infrastructure, architecture or API lock-in. In addition, the service control plane must support recursive service composition to be able to have end-to-end service visibility and control to avail the best resources wherever they are available to meet the quality of service dictated by business priorities, latency constraints and workload fluctuations. The application quality of service must not be dictated or limited by the infrastructure limitations. Then only we can predictably deploy highly reliable services on even not so reliable distributed infrastructure and increase efficiency to meet demand that is not as predictable.

Borrowing from biological and intelligent systems which specialize in exploiting  architectures that provide predictability, we can argue that infusing cognition into service management will provide such an architecture. Cognition [4] is associated with intent and its accomplishment through various processes that monitor and control a system and its environment. Cognition is associated with a sense of “self” (the observer) and the systems with which it interacts (the environment or the “observed”). Cognition [4] extensively uses time, history and reasoning in executing and regulating tasks that constitute a cognitive process. There is a fundamental reason why current Turing, von Neumann stored program computing model cannot address large-scale distributed computing with fluctuations both in resources and in computation workloads without increasing complexity and cost. As von Neumann [5] put it “It is a theorem of Gödel that the description of an object is one class type higher than the object.” An important implication of Gödel’s incompleteness theorem is that it is not possible to have a finite description with the description itself as the proper part. In other words, it is not possible to read yourself or process yourself as a process. In short, Gödel’s theorems prohibit “self-reflection” in Turing machines. Turing’s O-machine was designed to provide information that is not available in the computing algorithm executed by the TM. More recently, the super recursive algorithms proposed by Mark Burgin [3] points a way to model the knowledge about the hardware and software to reason and act to self-manage.  He proves that the super recursive algorithms are more efficient than plain Turing computations which assume unbounded resources.

Perhaps, we should look for “synthesis” solutions not in familiar places where we feel comfortable with more ad-hoc software and services that are labor and knowledge intensive. We should look for clues in biology, human organizational networks and even telecommunication networks to transform current datacenters from being infrastructure management systems to services switching centers of the future [6]. This requires search for new computing, management and programming models without disturbing current applications, operating systems or infrastructure while facilitating smooth migration to a more harmonious melody of orchestrated services on a global scale with high efficiency and resiliency.

References:

[1] Holbrook, Morris B. 2003. ” Adventures in Complexity: An Essay on Dynamic Open Complex Adaptive Systems, Butterfly Effects, Self-Organizing Order, Coevolution, the Ecological Perspective, Fitness Landscapes, Market Spaces, Emergent Beauty at the Edge of Chaos, and All That Jazz.” Academy of Marketing Science Review [Online] 2003 (6) Available: http://issuu.com/gfbertini/docs/adventures_in_complexity_-_an_essay_on_dynamic_ope/search 

[2] https://www.gartner.com/doc/2075838/infrastructure-service

[3] M. Burgin, Super-recursive Algorithms, New York: Springer, 2005.

[4] Mikkilineni, R. (2012). Applied Mathematics,  3, 1826-1835 doi:10.4236/am.2012.331248 Published Online November 2012 (http://www.SciRP.org/journal/am)

[5] Aspray W., and Burks A., 1987. Editors, Papers of John von Neumann on Computing and Computer Theory. In Charles Babbage Institute Reprint Series for the History of Computing, MIT Press. Cambridge, MA, p409, p.474.

[6] Rao Mikkilineni, “Designing a New Class of Distributed Systems” Springer, New York, 2011

Many-core Servers, Solid State Drives, and High-Bandwidth Networks: The Hardware Upheaval, The Software Shortfall, The Future of IT and all that Jazz
August 2, 2014

The “gap between the hardware and the software of a concrete computer and even greater gap between pure functioning of the computer and its utilization by a user, demands description of many other operations that lie beyond the scope of a computer program, but might be represented by a technology of computer functioning and utilization”

Marc S. Burgin, “Super Recursive Algorithms”, Springer, New York, 2005

Introduction

According to Holbrook (Holbrook 2003), “Specifically, creativity in all areas seems to follow a sort of dialectic in which some structure (a thesis or configuration) gives way to a departure (an antithesis or deviation) that is followed, in turn, by a reconciliation (a synthesis or integration that becomes the basis for further development of the dialectic). In the case of jazz, the structure would include the melodic contour of a piece, its harmonic pattern, or its meter…. The departure would consist of melodic variations, harmonic substitutions, or rhythmic liberties…. The reconciliation depends on the way that the musical departures or violations of expectations are integrated into an emergent structure that resolves deviation into a new regularity, chaos into a new order, surprise into a new pattern as the performance progresses.”

Current IT in this Jazz metaphor, evolved from a thesis and currently is experiencing an anti-thesis and is ripe for a synthesis that would blend the old and the new with a harmonious melody to create a new generation of highly scalable, distributed, secure services with desired availability, cost and performance characteristics to meet the changing business priorities, highly fluctuating workloads and latency constraints.

The Hardware Upheaval and the Software Shortfall

There are three major factors driving the datacenter traffic and their patterns:
1. A multi-tier architecture which determines the availability, reliability, performance, security and cost of initiating a user transaction to an end-point and delivering that service transaction to the user. The composition and management of the service transaction involves both the north-south traffic from end-user to the end-point (most often over the Internet) and the east-west traffic that flows through various service components such as DMZ servers, web servers, application servers and databases. Most often these components exist within the datacenter or connected through a WAN to other datacenters. Figure 1 shows a typical configuration.

Service Transaction Delivery Network

Service Transaction Delivery Network

The transformation from the client-server architectures to “composed service” model along with virtualization of servers allowing the mobility of Virtual Machines at run-time are introducing new patterns of traffic that increase in the east west direction inside the datacenter by orders of magnitude compared to the north-south traffic going from end-user to the service end-point or vice-versa. Traditional applications that evolved from client-server architectures use TCP/IP for all the traffic that goes across servers. While some optimizations attempt to improve performance for applications that go across servers using high-speed network technologies such as InfiniBand, Ethernet etc., TCP/IP and socket communications still dominate even among virtual servers within the same physical server.

2. The advent of many-core severs with tens and even hundreds of computing cores with high bandwidth communication among them drastically alters the traffic patterns.  When two applications are using two cores within a processor, the communication among them is not very efficient if it uses socket communication and TCP/IP protocol instead of shared memory. When the two applications are running in two processors within the same server, it is more efficient to use PCIExpress or other high-speed bus protocols instead of socket communication using TCP/IP. If the two applications are running in two servers within the same datacenter it is more efficient to use Ethernet or InfiniBand. With the advent of mobility of applications using containers or even Virtual Machines, it is more efficient to switch the communication mechanism based on the context of where they are running. This context sensitive switching is a better alternative to replicating current VLAN and socket communications inside the many-core server. It is important to recognize that the many-core servers and processors constitute a network where each node itself is a sub-network with different bandwidths and protocols (socket-based low-bandwidth communication between servers, InfiniBand, or PCI Express bus based communication across processors in the same server and shared memory based low latency communication across the cores inside the processor). Figure 2 shows the network of networks using many-core processors.

A Network of Networks with Multiple Protocols

A Network of Networks with Multiple Protocols

3. The many-core servers with new class of flash memory and high-bandwidth networks offer a new architecture for services creation, delivery and assurance going far beyond the current infrastructure-centric service management systems that have evolved from single-CPU and low-bandwidth origins. Figure 3 shows a potential architecture where many-core servers are connected with high-bandwidth networks that obviate the need for current complex web of infrastructure technologies and their management systems. The many-core servers each with huge solid-state Drives, SAS attached inexpensive disks, optical switching interfaces connected to WAN Routers offer a new class of services architecture if only the current software shortfall is plugged to match the hardware advances in server, network and storage devices.

If Server is the Cloud, What is the Service Delivery Network?

If Server is the Cloud, What is the Service Delivery Network?

This would eliminate the current complexity mainly involved in dealing with TCP/IP across east-west traffic and infrastructure based service delivery and management systems to assure availability, reliability, performance, cost and security. For example, current security mechanisms that have evolved from TCP/IP communications do not make sense across east/west traffic and emerging container based architectures with layer 7 switching and routing independent of server and network security offer new efficiencies and security compliance.

Current evolution of commodity clouds and distributed virtual datacenters while providing on-demand resource provisioning, auto-failover, auto-scaling and live-migration of Virtual machines, they are still tied to the IP address and associated complexity of dealing with infrastructure management in distributed environments to assure the end-to-end service transaction quality of service (QoS).

The QoS Gap

The QoS Gap

This introduces either vendor lock-in that precludes the advantages of commodity hardware or introduces complexity in dealing with multitude of distributed infrastructures and their management to tune the service transaction QoS. Figure 4 shows the current state of the art. One can quibble whether it includes every product available or whether they are depicted correctly to represent their functionality but the general picture describes the complexity and or vendor lock-in dilemma. The important point to recognize is that the service transaction QoS depends on tuning the SLAs of distributed resources at run-time across multiple infrastructure owners with disparate management systems and incentives. The QoS tuning of service transactions is not scalable without increasing cost and complexity if it depends on tuning the distributed infrastructure with a multitude of point solutions and myriad infrastructure management systems..

What the Enterprise IT Wants:

There are three business drivers that are at the heart of the Enterprise Quest for an IT framework:

  • Compression of Time-to-Market: Proliferation of mobile applications, social networking, and web-based communication, collaboration and commerce are increasing the pressure on enterprise IT to support a rapid service development, deployment and management processes. Consumer facing services are demanding quick response to rapidly changing workloads and the large-scale computing, network and storage infrastructure supporting service delivery requires rapid reconfiguration to meet the fluctuations in workloads and infrastructure reliability, availability, performance and security.
  • Compression of Time-to-Fix: With consumers demanding “always-on” services supporting choice, mobility and collaboration, the availability, performance and security of end to end service transaction is at a premium and IT is under great pressure to respond by compressing the time to fix the “service” regardless of which infrastructure is at fault. In essence, the business is demanding the deployment of reliable services on not so reliable distributed infrastructure.
  • Cost Reduction of IT operation and management which is running at about 60% to 70% of its budget going to keep the “service lights” on: Current service administration and management paradigm that originated with server-centric and low-bandwidth network architecture is resource-centric and assumes that the resources (CPU, memory, network bandwidth, latency, storage capacity, throughput and IOPs) allocated to an application at install time can be changed to meet rapidly changing workloads and business priorities in real-time. Current state-of-the art uses virtual servers, network and storage that can be dynamically provisioned using software API. Thus the application and service (a group of applications providing a service transaction) QoS (quality of service defining the availability, performance, security and cost) can be tuned by dynamically reconfiguring the infrastructure. There are three major issues with this approach:

With a heterogeneous, distributed and multi-vendor infrastructure, tuning the infrastructure requires myriad point solutions, tools and integration packages to monitor current utilization of the resources by the service components, correlate and reason to define the actions required and coordinate many distributed infrastructure management systems to reconfigure the resources.

In order to provide high availability and disaster recovery (HA/DR), recent attempts to move Virtual Machines (VM) introduces additional issues with IP mobility, Firewall reconfiguration, VM sprawl and associated run-away VM images, bandwidth and storage management.

Introduction of public clouds and the availability of software as a service, while they have worked well for new application development or non-mission critical applications or applications that can be re-architected to optimize for the Cloud API which leverage application/service components available, they are also adding additional cost for IT to migrate many existing mission critical applications that demand high security, performance and low-latency. The suggested Hybrid solutions require adopting new cloud architecture in the datacenters or use myriad orchestration packages that add additional complexity and tool fatigue.

In order to address the need to compress time to market and time to fix and to reduce the complexity, enterprises small and big are desperately looking for solutions.

The lines of business owners want:

  • End-to-end visibility and control of service QoS independent of who provides the infrastructure
  • Availability, performance and security governance based on policies
  • Accounting of resource utilization and dynamic resolution of contention for resources
  • Application architecture decoupled from infrastructure while still enabling continuous availability (or decouple functional requirements execution from non-functional requirement compliance)

IT wants to provide the application developers:

  • Application architecture decoupled from infrastructure by separating functional and non-functional requirements so that the application developers focus on business functions while deployment and operations are adjusted at run-time based on business priorities, latency constraints and workload fluctuations
  • Provide cloud-like services (on-demand provisioning of applications, self-repair, auto-scaling, live-migration and end-to-end security) at service level instead of at infrastructure level so that they can leverage own datacenter resources, or commodity resources abundant in public clouds without depending on cloud architectures, vendor API and cloud management systems.
  • Provide a suite of applications as a service (databases, queues, web servers etc.)
  • Service composition schemes that allow developers to reuse components and
  • Instrumentation to monitor service component QoS parameters (independent from infrastructure QoS parameters) to implement policy compliance
  • When problems occur provide component run-time QoS history to developers

IT wants to have the:

  • Ability to use local infrastructure or on demand cloud or managed out-sourced infrastructure
  • Ability to use secure cloud resources without cloud management API or cloud architecture dependence
  • Ability to provide end to end service level security independent of server and network security deployed to manage distributed resources
  • Ability to provide end-to-end service QoS visibility and control (on-demand service provisioning, auto-failover, auto-scaling, live migration and end-to-end security) across distributed physical or virtual servers in private or public infrastructure
  • Ability to reduce complexity and eliminate point solutions and myriad tools to manage distributed private and public infrastructure

Application Developers want:

  • To focus on developing service components, test them in their own environments and publish them in a service catalogue for reuse
  • Ability to compose services, test and deploy in their own environments and publish then in the service catalogue ready to deploy anywhere
  • Ability to specify the intent, context, constraints, communication, and control aspects of the service at run-time for managing non-functional requirements
  • An infrastructure that uses the specification to manage the run-time QoS with on-demand service provisioning on appropriate infrastructure (a physical or virtual server with appropriate service level assurance, SLA), manage run-time policies for fail-over, auto-scaling, live-migration and end-to-end security to meet run-time changes in business priorities, workloads and latency constraints.
  • Separation of run-time safety and survival of the service from sectionalizing, isolating, diagnosing and fixing at leisure
  • Get run-time history of service component behavior and ability to conduct correlated analysis to identify problems when they occur.

We need to discover a path to bridge the current IT to the new IT without changing applications, or the OSs or the current infrastructure while providing a way to migrate to a new IT where service transaction QoS management is truly decoupled from myriad distributed infrastructure management systems. This is not going to happen with current ad-hoc programming approaches. We need a new or at least an improved theory of computing.

As Cockshott et al (2012) point out current computing, management and programming models fall short when you try to include computers and the computed in same model.

“the key property of general-purpose computer is that they are general purpose. We can use them to deterministically model any physical system, of which they are not themselves a part, to an arbitrary degree of accuracy. Their logical limits arise when we try to get them to model a part of the world that includes themselves.”

There are emerging technologies that might just provide the synthesis (reconciliation depends on the way that the architecture departures or violations of expectations are integrated into an emergent structure that resolves deviation into a new regularity, chaos into a new order, surprise into a new pattern as the transformation progresses) required to build the harmony by infusing cognition into computing. Only future will tell if this architecture is expressive enough and efficient as Mark Burgin claims in his elegant book on “Super Recursive Algorithms” quoted above.

Is the Information Technology poised for a renaissance (with a synthesis) since the great masters (Turing, von Neumann, Shannon  etc.) developed the original thesis and take us beyond the current distributed-cloud-management anti-thesis.

The IEEE WETICE2015 International conference track on “the Convergence of Distributed Clouds, GRIDs and their Management” to be held in Cyprus next June (15 – 18) will address some of these emerging trends and attempt to bridge the old and the new.

24th IEEE International Conference on Enabling Technologies: Infrastructures for Collaborative Enterprises (WETICE2015)

24th IEEE International Conference on Enabling Technologies: Infrastructures for Collaborative Enterprises (WETICE2015) in Larnaca, Cyprus

 

For a state-of-the-emerging-science, please go to Infusing Self-Awareness into Turing Machine – A Path to Cognitive Distributed Computing presented in WETICE2014 in Parma Italy

 

 References:

Holbrook, Morris B. 2003. ” Adventures in Complexity: An Essay on Dynamic Open Complex Adaptive Systems, Butterfly Effects, Self-Organizing Order, Coevolution, the Ecological Perspective, Fitness Landscapes, Market Spaces, Emergent Beauty at the Edge of Chaos, and All That Jazz.” Academy of Marketing Science Review [Online] 2003 (6) Available: http://www.amsreview.org/articles/holbrook06-2003.pdf

Cockshott P., MacKenzie L. M., and  Michaelson, G, (2012) Computation and its Limits, Oxford University Press, Oxford.

Is Distributed IT Infrastructure Brokering (With or Without Virtual Machines) a Network Service that will be offered by a Network Carrier in the Future?
March 22, 2014

Trouble in IT Paradise with Darkening Clouds:

If you ask an enterprise CIO over a couple of drinks, what is his/her biggest hurdle today that is preventing  to deliver the business right resources at the right time at a right price, his/her answer would be that “the IT is too darn complex.” Over a long period of time, the infrastructure vendors have hijacked Information Technologies with their complex silos and expediency has given way to myriad tools and point solutions that overlay a management web. In addition, the Venture Capitalists looking for quickie “insertion points” with no overarching architectural framework have proliferated tools and appliances that have contributed to the current complexity and tool fatigue.

After a couple of more drinks, if you press the CIO why his/her mission critical applications are not migrating to the cloud which claims lesser complexity, the CIO laments that there is no cloud provider willing to sign a warranty that assures the service levels for their mission critical applications that guarantee application availability, performance and security levels. “Every cloud provider talks about infrastructure service levels but not willing to step up to assure application availability, performance and security.  There are myriad off-the main street providers that claim to offer orchestration to provide the service levels, but no one yet is signing on the dotted line.” The situation is more complicated when the resources span across multiple infrastructure providers.

The decoupling of the strong binding between the management of applications and the infrastructure management is a key for the CIO as more applications are developed with shorter time to market. CIO’s top five priorities are transnational applications demanding distributed resources, security, cost, compliance and uptime. A Gartner report claims that the CIOs spend 74% of IT budget on keeping the application “lights on” and another 18% on “changing the bulbs” and other maintenance activities. (It is interesting to recall that before Strowger’s switch eliminated many operators sitting in long rows plugging countless jacks  into countless plugs, the cost of adding and managing new subscribers was rising in a geometric proportion. According to the Bell System chronicles, one large city general manager of a telephone company at that time wrote that he could see the day coming soon when he would go broke merely by adding a few more subscribers because the cost of adding and managing a subscriber is far greater than the corresponding revenue generated. The only difference between today’s IT datacenter and central office before Strowger’s switch is that “very expensive consultants, countless hardware appliances, and countless software systems that manage them” replace “many operators, countless plugs and countless jacks”.)

In order to utilize commodity infrastructure while maintaining  high security, mobility for performance and availability, the CIOs are looking to solutions that let them focus on application quality of service (QoS) and are willing to outsource the infrastructure management to providers who can assure application mobility, availability and security albeit with end to end service visibility and control at their disposal.

While the public clouds seem to offer a way out to leverage the commodity infrastructure with on demand Virtual Machine provisioning, there are four hurdles that are preventing the CIO’s to embrace the clouds for mission critical applications:

  1. Current mission critical and even non-mission critical applications and services (groups of applications) are used to highly secure and low latency infrastructures that have been hardened and managed and the CIO’s are loath to spend more money to bring same level of SLA’s in public clouds.
  2. The dependence on particular service provider infrastructure API’s, Virtual Machine Image Management (nested or not) infrastructure dependencies and added self-healing, auto-scaling, live-migration service cost and complexity create service provider lock-in on their infrastructure and their management services. This defeats the intent to leverage the commodity infrastructure offered by different service providers.
  3. The increasing scope creep from infrastructure providers “up-the-stack” to provide application awareness and insert their API in application development in the name of satisfying non-functional requirements (availability, security, performance optimization) at run-time has started to increase the complexity and cost of application and service development. The resulting proliferation of tools and point solutions without a global architectural framework to use resources from multiple service providers have increased the integration and troubleshooting cost.
  4. Global communications, collaboration and commerce at the speed of light has increased the scale of computing and the distributed computing resource management has fallen short in meeting the scale and the fluctuations both caused by demand and also fluctuations in resources availability, performance and security.

The Inadequacy of Ad-hoc Programming to Solve Distributed Computing Complexity:

Unfortunately, the complexity is more a structural issue than an operational or infrastructure technology issue that cannot be resolved with ad-hoc programming techniques to manage the resources. Cockshott et al. conclude their book “Computation and its limits” with the paragraph “The key property of general-purpose computer is that they are general purpose. We can use them to deterministically model any physical system, of which they are not themselves a part, to an arbitrary degree of accuracy. Their logical limits arise when we try to get them to model a part of the world that includes themselves.”  While the success of IT in modeling and executing business processes has evolved to current distributed datacenters and cloud computing infrastructures that provide on-demand computing resources to model and execute business processes, the structure and fluctuations that dictate the evolution of computation have introduced complexity in dealing with real-time changes in the interaction of the infrastructure and the computations they perform. The complexity manifests in the following ways:

  1. In a distributed computing environment, maintaining the right computing resources (cpu, memory, network bandwidth, latency, storage capacity, throughput and IOPs) are available to right software component contributing to the service transaction requires orchestration and management of myriad computing infrastructures often owned by different providers with different profit motives and incentives. The resulting complexity in resource management to assure availability, performance and security of service transactions adds to the cost of computing. For example, it is estimated that up to 70% of current IT budget is consumed in assuring service availability, performance and security. The complexity is compounded in distributed computing environments that are supported by heterogeneous infrastructures with disparate management systems.
  2. In a large-scale dynamic distributed computation supported by myriad infrastructure components, the increased component failure probabilities introduce a non-determinism (for example the Google is observing emergent behavior in their scheduling of distributed computing resources when dealing with large number of resources) that must be addressed by a service control architecture that decouples functional and non-functional aspects of computing.
  3. Fluctuations in the computing resource requirements dictated by changing business priorities, workload variations that depend on service consumption profiles and real-time latency constraints dictated by the affinity of service components, all demand a run-time response to dynamically adjust the computing resources. Current dependence on myriad orchestrators and management systems cannot scale in a distributed infrastructure without either a vendor lock-in on infrastructure access methods or a universal standard that often stifles innovation and competition to meet fast changing business needs.

Thus the function, structure and fluctuations involved in dynamic processes delivering service transaction are driving a need to search new computation, management and programming models that address the unification of the computer and the computed and decouple the service management from the infrastructure management at run-time.

It is the Architecture Stupid:

A business process is defined both by functional requirements that dictate the business domain functions and logic as well as non-functional requirements that define operational constraints related to service availability, reliability, performance, security and cost dictated by business priorities, workload fluctuations and resource latency constraints. A non-functional requirement specifies criteria that can be used to judge the operation of a system, rather than specific behaviors. The plan for implementing functional requirements is detailed in the system design. The plan for implementing non-functional requirements is detailed in the system architecture. While much progress has been made in the system design and development, the architecture of distributed systems falls short to address the non-functional requirements for two reasons:

  1. Current distributed systems architecture from its server-centric and low-bandwidth origins has created layers of resource management-centric ad-hoc software to address various uncertainties that arise in a distributed environment. Lack of support for concurrency, synchronization, parallelism and mobility of applications dictated by the current serial von-Neumann stored program control has given rise to ad-hoc software layers that monitor and manage distributed resources. While this approach may have been adequate when distributed resources are owned by a single provider and controlled by a framework that provides architectural support for implementing non-functional requirements, the proliferation of commodity distributed resource clouds offered by different service providers with different management infrastructures adds scaling and complexity issues. Current OpenStack and AWS API discussions are a clear example that forces a choice of one or the other or increased complexity to use both.
  2. The resource-centric view of IT currently demotes application and service management to a second-class citizenship where the QoS of application/service is monitored and managed by myriad resource management systems overlaid with multiple correlation and analysis layers used to manipulate the distributed resources to adjust the Cpu, memory, bandwidth, latency, storage IOPs, throughput and capacity which are all what are required to keep the application/service to meet its quality of service. Obviously, this approach cannot scale unless single set of standards evolve or a single vendor lock-in occurs.

Unless an architectural framework evolves to decouple application/service management from myriad infrastructure management systems owned and operated by different service providers with different profit motives, the complexity and cost of management will only increase.

A Not So Cool Metaphor to Deliver Very Cool Services Anywhere, Anytime and On-demand:

A lesson on an architectural framework that addresses nonfunctional requirements while connecting billions of users anywhere anytime on demand is found in the Plain Old Telephone System (POTS). From the beginnings of AT&T to today’s remaking of at&t, much has changed but two things that remain constant are the universal service (on a global scale) and the telecom grade “trust” that are taken for granted. Very recently, Mark Zuckerberg proclaimed at the largest mobile technology conference in Barcelona that his very cool service Facebook wants to be the dial tone for the Internet. Originally, the dial tone was introduced to assure the telephone user that the exchange is functioning when the telephone is taken off-hook by breaking the silence (before an operator responded) with an audible tone.  Later on, the automated exchanges provided a benchmark for telecom grade trust that assures managed resources on-demand with high availability, performance and security.  Today, as soon as the user goes on hook, the network recognizes the profile based on the dialing telephone number.  As soon as the dialed party number is dialed, the network recognizes the destination profile and provisions all the network resources required to make the desired connection, commence billing, monitor and assure the connection till one of the parties initiates a disconnect. During the call, if the connection experiences any changes that impact the non-functional requirements, the network intelligence takes appropriate action based on policies. The resulting resiliency (availability, performance, and security), efficiency and scaling ability to connect billions of users on demand have come to be known as “Telecom grade trust”. An architectural flaw in the original service design (exploited by Steve Jobs by building a blue-box) was fixed by introducing an  architectural change to separate the data path and the control path. The resulting 800 service call model provided a new class of services such as call forwarding, call waiting and conference call.

The Internet on the other hand evolved to connect billions of computers together anywhere, anytime from the prophetic statement made by J. C. R. Licklider “A network of such (computers), connected to one another by wide-band communication lines [which provided] the functions of present-day libraries together with anticipated advances in information storage and retrieval and [other] symbiotic functions.”       The convergence of voice over IP, data and video networks has given rise to a new generation of services enabling communication, collaboration and commerce at the speed of light. The result is that the datacenter has replaced the central office to become the hub from which myriad voice, video and data services are created, and delivered on a global scale. However the management of these services which determines their resiliency, efficiency and scaling is another matter. In order to provide on demand services, anywhere, any-time with prescribed quality of service in an environment of wildly fluctuating workloads, changing business priorities and latency constraints dictated by the proximity of service consumers and suppliers, resources have to be managed in real-time across distributed pools to match the service QoS to resource SLAs. The telephone network is designed to share resources on a global scale and to connect them as required in real-time to meet the non-functional service requirements while current datacenters (whether privately owned or publicly provides as cloud services) are not. There are three structural deficiencies in the current distributed datacenter architecture to match the telecom grade resiliency, efficiency and scaling:

  1. The data path and service control path are not decoupled giving rise to same problems that Steve Jobs exploited causing a re-architecting of the network.
  2. The service management is strongly coupled with the resource management systems and does not scale as the resources become distributed and multiple service providers provide those resources with different profit motives and incentives. Since the resources are becoming commodity, every service provider wants to go up the stack to provide lock-in.
  3. Current trend to infuse resource management API in service logic to provide resource management at run-time and application aware architectures that want to establish intimacy with applications only increase complexity and make service composition with reusable service components all the more difficult because of their increased lock-in with resource management systems.

Resource management based datacenter operations miss an important feature of services/applications management which is that all services are not created equal. They have different latency and throughput requirements. They have different business priorities and different workload characteristics and fluctuations. What works for the goose does not work for the gander.  In addition to the current complexity and cost of resource management to assure service availability, reliability, performance and security, there is an even more fundamental issue that plagues the current distributed systems architecture. A distributed transaction that spans multiple servers, networks and storage devices in multiple geographies uses resources that span across multiple datacenters. The fault, configuration, accounting, performance and security (FCAPS) of a distributed transaction behavior requires the end-to-end connection management more like telecommunication service spanning distributed resources. Therefore, focusing on only resource management in a datacenter without the visibility and control of all resources participating in the transaction will not provide assurance of service availability, reliability, performance and security at run-time.

New Dial Tones for Application/Service Development, Deployment and Operation:

Current Web-scale applications are distributed transactions that span across multiple resources widely scattered across multiple locations owned and managed by different providers. In addition, the transactions are transient making connections with various components to fulfill an intent and closing them only to reconnect when they need them again. This is very much in contrast to always-on distributed computing paradigm of yesterday.

In creating, deploying and operating these services, there are three key stake holders and associated processes:

  1. Resources providers deliver the vital resources required to create, deploy and operate these resources on demand anywhere anytime (resource dial tone). The vital resources are just the CPU, memory, network latency, bandwidth and storage capacity, throughput and IOPs required to execute the application or service that has been compiled to “1”s and “0”s (the Turing Machine). The resource consumers care less about how you provide these as long as you maintain the service levels the resource providers agree to when the application or service requests the resources at provisioning time (matching the QoS request with SLA and maintaining it during the application/service life-time). The resource dial tone that assures the QoS with resource SLA is offered to two different types of consumers of this resource. First, the application developer who uses these resources to develop the service components and composes them to create more complex services with their own QoS requirements. Second the service operators who use the SLAs to provide management of QoS at run-time to deliver the services to end users.
  2. The application developers like to use their tools and best practices without any constraints from resource providers and the run-time vital signs required to execute their services should be transparent to where or who is providing the vital resources. The resources must support the QoS specified by developer or service composer depending on the context, communication, control and constraint needs. They do not care how they get the CPU, memory, bandwidth, storage capacity, throughput or IOPs or how the latency constraints are met. This model is a major departure from current SDN route focusing on giving control of resources to applications which is not a scalable solution that allows decoupling of resource management from service management.
  3. The service operators provide run-time QoS assurance by brokering the QoS demands to match the best available resource pool that meets the cost and quality constraints (the management dial tone that assures non-functional requirements). The brokering function is a network service ala services switching to match the applications/services to the right resources.

The brokering service must then provide the non-functional requirements management at run-time just as in POTS.

The New Service Operations Center (SOC) with End-to-end Service Visibility and Control Independent of Distributed Infrastructure Management Centers Owned by Different Infrastructure Providers:

The new Telco model that the broker facilitates allows the enterprises and other infrastructure users to focus on services architecture and management and use infrastructure as a commodity from different infrastructure providers just as Telcos provide shared resources with network services.

Telcograde

Figure 1: The Telco Grade Services Architecture that
decouples end to end service transaction management from infrastructure
management systems at run-time

The service broker matches the QoS of service and service components with service levels offered by different infrastructure providers based on the service blueprint which defines the context, constraints, communications and control abstraction of the service at hand. The service components are provided with desired Cpu, memory, bandwidth, latency, storage IOPs, throughput and capacity desired. The decoupling of service management from distributed infrastructure management systems puts the safety and survival of services first and allows sectionalization, isolation, diagnosis anfd fixing infrastructure at leisure as is the case today with POTS.

It is important to note that the service dial tone Zuckerberg is talking about is not related to the resources dial tone or management dial tone required for providing service connections and management at run-time. He is talking about application end user receiving the content. Facebook application developers do not care how the computing resources are provided as long as their service QoS is maintained to meet the business priorities, workloads and latency constraints to deliver their service on a global scale. Facebook CIO would rather spend time maintaining the service QoS by getting the resources wherever they are available to meet the service needs at reasonable cost. In fact most CIOs would get rid of the infrastructure management burden if they have QoS assurance and end-to-end service visibility and service control (they could not care less about access to resources or their management systems) to manage the non-functional requirements at run-time. After all, Facebook’s open compute project is a side effect trying to fill a gap left by infrastructure providers – not their main line of business. The crash that resulted after Zuckerberg’s announcement of WhatsApp acquisition was not the “cool” application’s fault. They probably could have used a service broker/switch providing the old fashioned resource dial tone so that they could provide the service dial tone to their users.

This is similar to a telephone company assuring appropriate resources to connect different users based on their profiles or the Internet connecting devices based on their QoS needs at run-time. The broker acts as service switch that connects various service components at run-time and matches the QoS demands with appropriate resources.

With the right technology, the service broker/switch may yet provide the required service level warranties to the enterprise CEOs from well-established carriers with money and muscle.

Will at&t and other Telcos have the last laugh by incorporating this brokering service switch in the network and make current distributed datacenters (cloud or otherwise with physical or virtual infrastructure) a true commodity?

A Path Toward Intelligent Services using Dumb Infrastructure on Stupid, Fat Networks?
November 3, 2013

“The key property of general-purpose computer is that they are general purpose. We can use them to deterministically model any physical system, of which they are not themselves a part, to an arbitrary degree of accuracy. Their logical limits arise when we try to get them to model a part of the world that includes themselves.”

—  P. Cockshott, L. M. MacKenzie and  G. Michaelson, Computation and its Limits, Oxford University Press, Oxford 2012, p 215

“The test of a framework, however, is not what one can say about it in principle but what one can do with it”

—  Andrew Wells, Rethinking Cognitive Computation: Turing and the Science of the Mind, Palgrave, Macmillan, 2006.

Summary

The “Convergence of Clouds, Grids and their Management” conference track is devoted to discussing current and emerging trends in virtualization, cloud computing, high-performance computing, Grid computing and cognitive computing. The tradition that started in WETICE2009 “to analyze current trends in Cloud Computing and identify long-term research themes and facilitate collaboration in future research in the field that will ultimately enable global advancements in the field that are not dictated or driven by the prototypical short term profit driven motives of a particular corporate entity” has resulted in a new computing model that was included in the Turing Centenary Conference proceedings in 2012. More recently, a product based on these ideas was discussed in the 2013 Open Server Summit (www.serverdesignsummit.com), where many new ideas and technologies were presented to exploit the new generation of many-core servers, high-bandwidth networks and high-performance storage. We present here some thoughts on current trends which we hope will stimulate further research to be discussed in the WETICE 2014 conference track in Parma, Italy (http://wetice.org).

Introduction

Current IT datacenters have evolved from their server-centric, low-bandwidth origins to distributed and high-bandwidth environments where resources can be dynamically allocated to applications using computing, network and storage resource virtualization. While Virtual machines improve resiliency and provide live migration to reduce the recovery time objectives in case of service failures, the increased complexity of hypervisors, their orchestration, Virtual Machine images and their movement and management adds an additional burden in the datacenter.

Further automation trends continue to move toward static applications (locked-in-a-virtual machine, often as one application in one virtual machine) in a dynamic infrastructure (virtual servers, virtual networks, virtual storage, Virtual Image managers etc.). The safety and survival of applications and end to end service transactions delivered by a group of applications are managed by dynamically monitoring and controlling the resources at run-time in real-time. As services migrate to distributed environments where applications contributing to a service transaction are deployed in different datacenters and public or private clouds often owned by different providers, resource management across distributed resources is provided using myriad point solutions and tools that monitor, orchestrate and control these resources. A new call for application-centric infrastructure proposes that the infrastructure provide (http://blogs.cisco.com/news/application-centric-infrastructure-a-new-era-in-the-data-center/ ):

  • Application Velocity (Any workload, anywhere): Reducing application deployment time through a fully automated and programmatic infrastructure for provisioning and placement. Customers will be able to define the infrastructure requirements of the application, and then have those requirements applied automatically throughout the infrastructure.
  • A common platform for managing physical, virtual and cloud infrastructure: The complete integration across physical and virtual, normalizing endpoint access while delivering the flexibility of software and the performance, scale and visibility of hardware across multi-vendor, virtualized, bare metal, distributed scale out and cloud applications
  • Systems Architecture: A holistic approach with the integration of infrastructure, services and security along with the ability to deliver simplification of the infrastructure, integration of existing and future services with real time telemetry system wide.
  • Common Policy, Management and Operations for Network, Security, Applications: A common policy management framework and operational model driving automation across Network, Security and Application IT teams that is extensible to compute and storage in the future.
  • Open APIs, Open Source and Multivendor: A broad ecosystem of partners who will be empowered by a comprehensive published set of APIs and innovations contributed to open source.
  • The best of Custom and Merchant Silicon: To provide highly scalable, programmatic performance, low-power platforms and optics innovations that protect investments in existing cabling plants, and optimize capital and operational expenditures.

Perhaps this approach will work in a utopian IT landscape where either the infrastructure is provided by a single vendor or universal standards force all infrastructures to support common API. Unfortunately the real world evolves in a diverse, heterogeneous and competitive environment and what we are left with is a strategy that cannot scale and lacks end-to-end service visibility and control. End-to-end security becomes difficult to assure because of the myriad security management systems that control distributed resources. The result is open source systems that attempt to fill this niche. Unfortunately, in a highly networked world where multiple infrastructure providers provide a plethora of diverse technologies that evolve at a rapid rate to absorb high-paced innovations, orchestrating the infrastructure to meet the changing workload requirements that applications must deliver is a losing battle. The complexity and tool fatigue resulting from layers of virtualization and orchestration of orchestrators is crippling the operation and management of datacenters (virtualized or not) requiring 70% of current IT budgets going toward keeping the lights on. An explosion of tools, special purpose appliances (for Disaster Recovery, IP security, Performance optimization etc.) and administrative controls have escalated operation and management costs. Gartner Report estimates that for every 1$ spent on development of an application, another $1.31 is spent on assuring safety & survival. While all vendors agree upon Open Source, Open API, and multi-vendor support, reality is far from it. An example is the recent debate about whether OpenStack should include Amazon AWS API support while the leading cloud provider conveniently ignores the competing API.

The Strategy of Dynamic Virtual Infrastructure

The following picture presented in the Open Server Summit Presents a vision of future datacenter with a virtual switch network overlay over physical network.

Virtual NetworkFigure 1: Network Virtualization: What It Is and Why It Matters – Presented in Open Server Summit in 2013

Bruce Davie, Principal Engineer, VMware

In addition to the Physical network connecting physical servers, an overlay of virtual network inside the physical server to connect the virtual machines inside a physical server. In addition, a plethora of virtual machines are being introduced to replace the physical routers and switches that control the physical network. The quest to dynamically reconfigure the network at run-time to meet the changing application workloads, business priorities and latency constraints has introduced layers of additional network infrastructure albeit software-defined. While applications are locked in a virtual server, the infrastructure is evolving to dynamically reconfigure itself to meet changing application needs. Unfortunately this strategy can not scale in a distributed environment where different infrastructure providers deploy myriad heterogeneous technologies and management strategies and results in orchestrators of orchestrators contributing to complexity and tool fatigue in both datacenters and clod environments (private or public).

Figure 2 shows a new storage management architecture also presented in the Open Server Summit.

Virtual Storage

Figure 2: PCI Express Supersedes SAS and SATA in Storage – Presented in Open Server Summit 2013, Akber Kazmi, Senior Marketing Director, PLX Technology

The PCIe switch allows a converged physical storage fabric at half the cost and half the power of current infrastructure. In order to leverage these benefits, the management infrastructure has to accommodate it which adds to the complexity.

In addition, it is estimated that the data traffic inside the datacenter is about 1000 times that of the data that is sent to and received from the users outside. This completely changes the role of TCP/IP traffic inside the datacenter and consequently the communication architecture between applications inside the datacenter. It does not anymore make sense for Virtual machines running inside a Many-core server to use TCP/IP as long as they are within the datacenter. In fact, it makes more sense for them to communicate via shared memory when they are executed on different cores within a processor, communicate via high speed bus when they are executed on different processors in the same server and a high speed network when they are executed in different servers in the same datacenter. TCP/IP is only needed when communicating with users outside the datacenter who can only be accessed via the Internet.

Figure 3 shows the server evolution.

Server

Figure 3: Servers for the New Style of IT – Presented in Open Server summit 2013, Dwight Barron, HP Fellow and Chief Technologies Hyper-scale Server Business Segment, HP Servers Global Business Unit, Hewlett-Packard

As the following picture presents, current evolution of the datacenter is designed to provide dynamic control of resources for addressing the work-load fluctuations at run-time, changing business priorities and real-time latency constraints. The applications are static in a Virtual or Physical Server and the software defined infrastructure dynamically adjusts to changing application needs.

Complexity1

Figure 4: Macro Trends, Complexity, and SDN – Presentation in the Open Server Summit 2013, David Meyer, CTO/Chief Architect, Brocade

Cognitive Containers & Self-Managing Intelligent Services on Static Infrastructure

With the advent of many-core servers, high bandwidth technologies connecting these servers, and new class of high performance storage devices that can be optimized to meet the workload needs (IOPs intensive, throughput sensitive or capacity hungry), is it time to look at a static infrastructure with dynamic application/service management to reduce IT complexity in both datacenters and clouds (public or private)? This is possible if we can virtualize the applications inside a server (physical or virtual) and decouple the safety and survival of the applications and groups of applications that contribute to a distributed transaction from myriad resource management systems that provision and control a plethora of distributed resources supporting these applications.

The Cognitive Container discussed in the Open Server Summit (http://lnkd.in/b7-rfuK) presents the decoupling required between application and service management and underlying distributed resource management systems. Cognitive Container is specially designed to decouple the management of an application and service transactions that a group of distributed applications execute from the infrastructure management systems, at run-time, controlling their resources that are often owned or operated by different providers. The safety and survival of the application at run-time is put ahead by infusing the knowledge about the application (such as the intent, non-functional attributes, run-time constraints, connections and communication behaviors) into the container and using this information to monitor and manage the application at run-time. The Cognitive Container is instantiated and managed by a Distributed Cognitive Transaction Platform (DCTP) that sits between the applications and the OS facilitating the run-time management of Cognitive Containers. The DCTP does not require any changes to the application, OS or the infrastructure and uses the local OS in a physical or virtual server. A network of Cognitive Containers infused with similar knowledge about the service transaction they execute also is managed at run-time to assure the safety and survival based on policies dictated by business priorities, run-time workload fluctuations and real-time latency constraints. The Cognitive Container network using replication, repair, recombination and reconfiguration properties provide dynamic service management independent of infrastructure management systems at run-time. The Cognitive Containers are designed to use the local operating system to monitor the application vital signs (CPU, memory, bandwidth, latency, storage capacity, IOPs and throughput) and run-time behavior to manage the application to conform to the policies.

The cognitive container can be deployed in a physical or virtual server and does not require any changes to the applications, OSs or the infrastructure. Only the knowledge about the functional and n0n-functional requirements has to be infused into the Cognitive Container. The following figure shows a Cognitive Network deployed in a distributed infrastructure. The Cognitive Container and the service management are designed to provide auto-scaling, self-repair, live-migration and end-to-end service transaction security independent of infrastructure management system.

service visibility

Figure 5: End-to-End Service Visibility and Control in a Distributed Datacenter (Virtualized or Not) – Presented in the Open Server Summit

Rao Mikkilineni, Chief Scientist, C3 DNA

Using the Cognitive Container network it is possible to create a federated service creation, delivery and assurance platforms that transcend the physical and virtual server boundaries and geographical locations as shown in figure below.

 Platform

Figure 6: Federated Services Fabric with Service Creation, delivery and assurance processes decoupled from Resource provisioning, management and control.

This architecture provides an opportunity to simplify the infrastructure where a tiered server, storage and network infrastructure that is static and hardwired to provide various servers (physical or virtual) with specified service levels (CPU, memory, network bandwidth, latency, storage capacity and throughput) the cognitive containers are looking for based on their QoS requirements. It does not matter what technology is used to provision these servers with required service levels. The Cognitive Containers monitor these vital signs using the local OS and if they are not adequate, they will migrate to other servers where they are adequate based on policies determined by business priorities, run-time workload fluctuations and real-time latency constraints.

The infrastructure provisioning then becomes a simple matter of matching the Cognitive Container to the server based on QoS requirements. Thus the Cognitive Container services network provides a mechanism to deploy intelligent (self-aware, self-reasoning and self-controlling) services using dumb infrastructure with limited intelligence about services and applications (matching application profile to the server profile) on stupid pipes that are designed to provide appropriate performance based on different technologies as discussed in the Open Server Summit.

The managing and safekeeping of application required to cope with a non-deterministic impact on workloads from changing demands, business priorities, latency constraints, limited resources and security threats is very similar to how cellular organisms manage life in a changing environment. The managing and safekeeping of life efficiently at the lowest level of biological architecture that provides the resiliency was in his mind when von Neumann was presenting his Hixon lecture (Von Neumann, J. (1987) Papers of John von Neumann on Computing and Computing Theory, Hixon Symposium, September 20, 1948, Pasadena, CA, The MIT Press, Massachusetts, p474). ‘‘The basic principle of dealing with malfunctions in nature is to make their effect as unimportant as possible and to apply correctives, if they are necessary at all, at leisure. In our dealings with artificial automata, on the other hand, we require an immediate diagnosis. Therefore, we are trying to arrange the automata in such a manner that errors will become as conspicuous as possible, and intervention and correction follow immediately.’’ Comparing the computing machines and living organisms, he points out that the computing machines are not as fault tolerant as the living organisms. He goes on to say ‘‘It’s very likely that on the basis of philosophy that every error has to be caught, explained, and corrected, a system of the complexity of the living organism would not run for a millisecond.’’ Perhaps the Cognitive Container bridges this gap by infusing self-management into computing machines that manage the external world while also managing themselves with self-awareness, reasoning, and control based on policies and best practices.

Cognitive Containers or not, the question is how do we address the problem of ever increasing complexity and cost in current datacenter and cloud offerings? This will be a major theme in the 4th conference track on the Convergence of Distributed Clouds, Grids and their management at WETICE2014 in Parma, Italy.

WETICE 2014 and The Conference Track on the Convergence of Clouds, Grids and Their Management
October 30, 2013

WETICE is an annual IEEE International conference on state-of-the-art research in enabling technologies for collaboration, consisting of a number of cognate conference tracks. The “Convergence of Clouds, Grids and their Management” conference track is devoted to discussing current and emerging trends in virtualization, cloud computing, high performance computing, Grid computing and Cognitive Computing. The tradition that started in WETICE2009 “to analyze current trends in Cloud Computing and identify long-term research themes and facilitate collaboration in future research in the field that will ultimately enable global advancements in the field that are not dictated or driven by the prototypical short term profit driven motives of a particular corporate entity” has resulted in a new computing model that was included in the Turing Centenary Conference proceedings in 2012. The 2013 conference track discussed Virtualization, Cloud Computing and the Emerging Datacenter Complexity Cliff in addition to conventional cloud and grid computing solutions.

The WETICE 2014 conference to be held in Parma, Italy during June, 23rd-25th, 2014, will continue the tradition by continuing the discussions on the convergence of clouds, grids and their management. In addition, it will also solicit papers on new computing models, cognitive computing platforms and strong AI resulting from recent efforts to inject cognition into computing (Turing Machines).

All papers are refereed by the Scientific Review of Committee of each conference track. All accepted papers will be published in the electronic proceedings by the IEEE Computer Society, and submitted to the IEEE digital library. The proceedings will be submitted for indexing through INSPEC, Compendex, and Thomson Reuters, DBLP, Google Scholar and EI Index.

http://wetice.org

Cellular Resiliency, Turing Machines, New Computing Models and the Zen of Consciousness
June 23, 2013

“WETICE 2012 Convergence of Distributed Clouds, Grids and their Management Conference Track is devoted to transform current labor intensive, software/shelf-ware-heavy, and knowledge-professional-services dependent IT management into self-configuring, self-monitoring, self-protecting, self-healing and self-optimizing distributed workflow implementations with end-to-end resource management by facilitating the development of a Unified Theory of Computing.”

Here is more food for thought…

Abstract:

Cellular biology has evolved to capture dynamic representations of self and its surroundings and a systemic view of monitoring and control of both the self and the surroundings to optimize the organism’s chances of survival. Signaling plays a key role in shaping the structure and behavior of cellular organisms to exhibit a high degree of resiliency by monitoring and controlling its own activity and its interactions with the outside environment with a Zen-like one-ness of the observer and the observed. Evolution has invented the genetic transactions of replication, repair, recombination and reconfiguration to support the survival of living cells by organizing themselves to execute a coordinated set of activities and signaling provides a vehicle for managing the system-wide behavior.

By introducing signaling and self-management in a Turing node and a signaling network as an overlay over the computing network, the current von-Neumann computing model is evolved to bring the architectural resiliency of cellular organisms to computing infrastructure. The new approach introduces the genetic transactions of replication, repair, recombination and reconfiguration to program self-resiliency in distributed computing systems executing a managed workflow. Perhaps, the injection of parallelism and network based composition of “Self” identity are the first steps in introducing the elements of homeostasis and self-management required for developing consciousness in the computing infrastructure.

Introduction:

As recent advances in neuroscience throw new light on the process of evolution of the cellular computing models, it is becoming clear that communication and collaboration mechanisms of distributed computing elements and end-to-end distributed transaction management played a crucial role in the development of self-resiliency, efficiency and scaling which are exhibited by diverse forms of life from the cellular organisms to highly evolved human beings. According to Antonio Damasio (Damasio 2010), managing and safe keeping life is the fundamental premise of biological value and this biological value has influenced the evolution of brain structures. “Life regulation, a dynamic process known as homeostasis for short, begins in unicellular living creatures, such as bacterial cell or a simple amoeba, which do not have a brain but are capable of adaptive behavior. It progresses in individuals whose behavior is managed by simple brains, as in the case with worms, and it continues its march in individuals whose brains generate both behavior and mind (insects and fish being examples)….” Homeostasis is the property of a system that regulates its internal environment and tends to maintain a stable, constant condition of properties like temperature or chemical parameters that are essential to its survival. System-wide homeostasis goals are accomplished through a representation of current state, desired state, a comparison process and control mechanisms.

He goes on to say that “consciousness came into being because of biological value, as a contributor to more effective value management. But consciousness did not invent biological value or the process of valuation. Eventually, in human minds, consciousness revealed biological value and allowed the development of new ways and means of managing it.” The governance of life’s processes is present even in single-celled organisms that lack a brain and it has evolved to the conscious awareness which is the hallmark of highly evolved human behavior. “Deprived of conscious knowledge, deprived of access to the byzantine devices of deliberation available in our brains, the single cell seems to have an attitude: it wants to live out its prescribed genetic allowance. Strange as it may seem, the want, and all that is necessary to implement it, precedes the explicit knowledge and deliberation regarding life conditions, since the cell clearly has neither.  The nucleus and the cytoplasm interact and carry out complex computations aimed at keeping the cell alive.  They deal with the moment-to-moment problems posed by the living conditions and adapt the cell to the situation in a survivable manner. Depending on the environmental conditions, they rearrange the position and distribution of molecules in their interior, and they change the shape of sub-components, such as microtubules, in an astounding display of precision.  They respond under duress and under nice treatment too. Obviously, the cell components carrying out those adaptive adjustments were put into place and instructed by the cell’s genetic material.”  This vivid insight brings to light the cellular computing model that:

  1. Spells out the computational workflow components as a stable sequence of patterns that accomplishes a specific purpose,
  2. Implements a parallel management workflow with another sequence of patterns that assures the successful execution of the system’s purpose (the computing network to assure biological value with management and  safekeeping),
  3. Uses a signaling mechanism that controls the execution of the workflow for gene expression (the regulatory network) and
  4. Assures real-time monitoring and control (homeostasis) to execute genetic transactions of replication, repair, recombination and reconfiguration (Stanier, Moore, 2006).

The managing and safekeeping life efficiently are evident at the lowest level of biological architecture that provides the resiliency that von Neumann was discussing in his Hixon lecture (von Neumann, 1987). ‘‘The basic principle of dealing with malfunctions in nature is to make their effect as unimportant as possible and to apply correctives, if they are necessary at all, at leisure. In our dealings with artificial automata, on the other hand, we require an immediate diagnosis. Therefore, we are trying to arrange the automata in such a manner that errors will become as conspicuous as possible, and intervention and correction follow immediately.’’ Comparing the computing machines and living organisms, he points out that the computing machines are not as fault tolerant as the living organisms. He goes on to say ‘‘It’s very likely that on the basis of philosophy that every error has to be caught, explained, and corrected, a system of the complexity of the living organism would not run for a millisecond.’’

The connection between consciousness and computing models is succinctly summarized by Samad and Cofer (Samad, Cofer, 2001).  While there is no accepted precise definition of the term consciousness, “it is generally held that it is a key to human (and possibly other animal) behavior and to the subjective sense of being human. Consequently, any attempt to design automation systems with humanlike autonomous characteristics requires designing in some elements of consciousness.  In particular, the property of being aware of one’s multiple tasks and goals within a dynamic environment and of adapting behavior accordingly.” They point to two theoretical limitations of formal systems that may inhibit the implementation of computational consciousness and hence limit our ability to design human-like autonomous systems. “First, we know that all digital computing machines are “Turing-equivalent”-They differ in processing speeds, implementation technology, input/output media, etc., but they are all  (given unlimited memory and computing time) capable of exactly the same calculations. More importantly, there are some problems that no digital computer can solve. The best known example is the halting problem; we know that it is impossible to realize a computer program that will take as input another, arbitrary, computer program and determine whether or not the program is guaranteed to always terminate.

Second, by Gödel’s proof, we know that in any mathematical system of at least a minimal power there are truths that cannot be proven. The fact that we humans can demonstrate the incompleteness of a mathematical system has led to the claims that Gödel’s proof does not apply to humans.”

An important implication of Gödel’s incompleteness theorem is that it is not possible to have a finite description with the description itself as the proper part. In other words, it is not possible to read yourself or process yourself as process. In short, Gödel’s theorems prohibit “self-reflection” in Turing machines. Louis Barrett highlights (Barrett, 2011) the difference between Turing Machines implemented using von Neumann architecture and biological systems. “Although the computer analogy built on von Neumann architecture has been useful in a number of ways, and there is also no doubt that work in classic artificial intelligence (or, as it is often known, Good Old Fashioned AI: GOFAI) has had its successes, these have been somewhat limited, at least from our perspective here as students of cognitive evolution.” She argues that the Turing machines based on algorithmic symbolic manipulation using von Neumann architecture, gravitate toward those aspects of cognition, like natural language, formal reasoning, planning, mathematics and playing chess, in which the processing of abstract symbols in a logical fashion and leaves out other aspects of cognition that deal with producing adoptive behavior in a changeable environment. Unlike the approach where perception, cognition and action are clearly separated, she suggests that the dynamic coupling between various elements of the system, where each change in one element continually influences every other element’s direction of change has to be accounted for in any computational model that includes system’s sensory and motor functions along with analysis. To be fair, such couplings in the observed can be modeled and managed using a Turing machine network and the Turing network itself can be managed and controlled by another serial Turing network.  What is not possible is the tight integration of the models of the observer/manager and the observed/managed with a description of the “self” (or a specification of the manager) using parallelism and signaling that are the norm and not an exception in biology.

A more interesting controversy that has erupted regarding the need for new computing models (Wegner, Eberbach, 2004, Cockshott, Michaelson, 2007, Goldin, Wegner, 2008) throws some new light on the need for re-examining the Turing machines, Gödel’s prohibition of self-reflection  and von Neumann’s conjecture. An even more recent discussion of the need for new computing models was presented in the Ubiquity symposium (ACM Ubiquity, 2011). As we describe later, these authors are attempting to address how to model computational problems that cannot be solved by a single Turing machine but can be solved using a set of Turing machines interacting with each other. In particular, the property of being aware of one’s multiple tasks and goals within a dynamic environment and of adapting behavior accordingly which is related to consciousness mentioned earlier is one such problem that a single Turing machine can not solve. The insights into biology suggest that in order to model temporal dynamics of the observer and the observed while also assuring the safe-keeping of the observer (with a “self” identity) requires modifications to the Turing machine to accommodate changes to the behavior while computation is still in progress.

Self, Consciousness, and Emotions – The Dynamic Representation of the Observer and the Observed:

Self-reflection, setting expectations, monitoring the deviations and taking corrective action are essential for managing the business of life through homeostasis and evolution has figured out how to encapsulate the right descriptions to execute the life’s processes using the genetic transaction of replication, repair, recombination and reconfiguration by exploiting parallelism and signaling. As Jonah Lehrer (Lehrer, 2010) describes in his book “How We Decide”, “Dopamine neurons automatically detect the subtle patterns that we would otherwise fail to notice; they assimilate all the data that we can’t consciously comprehend. And, then, once they come up with a set of refined predictions about how the world works, they translate these predictions to emotions.” Emotions, it seems are the instinctual localized component level suggestions for corrective actions based on local experience. Conscience [1] on the other hand, is the adult who correlates the instinctual suggestions with much larger perspective and makes decisions based on global priorities.

It is becoming clear from the recent advances in neuroscience, that self-reflection is a key component in living organisms.  Homeostasis is not possible without a dynamic and active representation of the observer and the observed.

A cellular organism is the simplest form of life that maintains an internal environment that supports its essential biochemical reactions, despite changes in the external environment. Therefore, a selectively permeable plasma membrane surrounding a concentrated aqueous solution of chemicals is a feature of all cells. In addition it is capable of self-replication and self-repair which may be unicellular or multicellular. Unicellular organisms perform all the functions of life. Multicellular organisms contain several different cell types that are specialized to perform specific functions. The cell adapts to its environment by recognition and transduction of a broad range of environmental signals, which in turn activate response mechanisms by regulating the expression of proteins that take part in the corresponding processes. The nucleus of the cell houses deoxyribonucleic acid (DNA) the genetic blueprint of the organism which determines the structure and function of the organism as a whole. The DNA serves two functions. First, it contains instructions for assembling the structural and enzymatic proteins of the cell. Cellular enzymes in turn control the formation of other cellular structures and also determine the functional activity of the cell by regulating the rate at which metabolic reactions proceed. Second, by replicating (making copies of itself), DNA perpetuates the genetic blueprint within all new cells formed within the body and is responsible for passing on genetic information from the survivors to successors.

A gene is a stretch of DNA that contains instructions or code for a particular function such as synthesizing a protein or dictating the assembly of amino acids. A unique set of genes are packaged as chromosomes in complex organisms. A gene regulatory network represents relationships between genes that can be established from measuring how the expression level of each one affects the expression level of the others. In any global cellular network, genes do not interact directly with other genes. Instead, gene induction or repression occurs, the action of specific proteins, which are in turn products of certain genes as well. In essence, gene networks are abstract models that display causal relationships between gene activities and are represented by directed graphs. Nearly all of the cells of a multicellular organism contain same DNA. Yet this same genetic information yields a large number of different cell types. The fundamental difference between a neuron and a liver cell, for example, is which genes are expressed. The regulatory gene network forms a cellular control circuitry defining the overall behavior of the various cells. According to Antonio Damasio (Damasio, 2010), the brain architecture is an evolutionary aid to the business of managing life which consists of managing the body and the management gains precision and efficiency with the presence of circuits of neurons assisting the management. In describing the role of neurons, he says that “neurons are about life and managing life in other cells of the body, and that that aboutness requires two-way signaling. Neurons act on other body cells, via chemical messages or excitation of muscles, but in order to do their job, they need inspiration from the very body they supposed to prompt, so to speak. In simple brains, the body does its prompts simply by signaling to subcortical nuclei. Nuclei are filled with “dispositional know-how,” the sort of knowledge that does not require detailed mapped representations. But in complex brains, the map-making cerebral cortices describe the body and its doings in so much explicit detail that the owners of those brains become capable, for example, of “imaging: the shape of their limbs and their positions in space, or the fact that their elbows hurt or their stomach does”.

The complex network of neural connections and signaling mechanisms collaborate to create a dynamic, active and temporal representation of both the observer and the observed with myriad patterns, associations and constraints among their components. It seems that the business of managing life is more than mere book-keeping that is possible with a Turing machine. It involves the orchestration of an ensemble with a self-identity both at the group and the component level contributing to the system’s biological value. It is a hierarchy of individual components where each node itself is a sub-network with its own identity and purpose which is consistent with the system-wide purpose. To be sure, each component is capable of book-keeping and algorithmic manipulation of symbols. In addition, identity and representations of the observer and the observed at both the component and group level make system-wide self-reflection possible.

In short, the business of managing life is implemented by a system consisting of a network of networks with multiple parallel links that transmit both control information and the mission critical data required to sense and to control the observed by the observer. The data and control networks provide the capabilities to develop an internal representation of both the observer and the observed along with the processes required to implement the business of managing life. The organism is made up of autonomic components making up an ensemble collaborating and coordinating a complex set of life’s processes that are executed to sense and control both the observer and the observed.  In this sense, the brain and the body are part of a collaborating system that has a unique identity and a structure that preserves the interrelationships.  The system consists of:

  1. Components each with a purpose within a larger system (specialization)
  2. All of a component parts must be present for the system to carry out its purpose optimally,
  3. A system’s parts must be arranged in a specific way for the system to carry out its purpose (separation of concerns),
  4. Systems change in response to feedback (collect information, analyze information and control environment using specialized resources), and
  5. Systems maintain their stability (in accomplishing their purpose) by making adjustments based on feedback (homeostasis).

[1] According to Antonio Damasio (Damasio, 2010), consciousness pertains to the knowing of any object or action attributed to a self, while conscience pertains to the good or evil to be found in actions or objects. The identity of self and its safekeeping are essential parts of life processes.  “The non-conscious neural signaling of an individual organism begets the proto-self which permits core self and core consciousness, which allow for an auto-biographical self which permits extended consciousness. At the end of the chain, extended consciousness permits conscience.”

Figure 1 shows the model of core-conscience, its relationship to the Observed and the extended conscience (Damasio, 1999) proposed by Damasio based on his studies in neuroscience.

Figure 1: The mapping of the observer, the observed and myriad models, associations and processes executed using parallel signaling and data exchange networks.  Each component itself is a sub-network with a purpose defined by its own internal models.

Literature is filled with discussion about Gödel’s prohibition of self-reflection in Turing machines and why consciousness cannot emerge from the brain models that depend on Turing machines.  There are many theories on how the human brain is unique and may even involve quantum phenomena or gravity waves (Scott, 1995 and Davis 1992).  However Damasio (Damasio, 2010) takes the evolutionary approach to discuss genomic unconsciousness, the feeling of conscious will, educating the cognitive conscious, the reflective self and its consequences. He goes on to say “in one form or another, the cultural developments manifest the same goal as the form of automated homeostasis.” “They respond to a detection of the imbalance in the life process, and seek to correct it within the constraints of human biology and of the physical and social environment.”

Instead of adding to the already existing controversy (Scott, 1995) on consciousness, we take a different route using Damasio’s emphasis on homeostasis along with the dynamic representation of the observer and the observed. We apply them to extend the Turing machine and its von Neumann Serial computing implementation.  We ask how we can utilize the abstractions that assist in the business of managing life in cellular organisms, discussed above, to enhance the resiliency of distributed computing systems. In the next section we analyze the current implementation of Turing machines and suggest adding some of the abstractions that have proven useful in managing life’s processes to develop a computing model that addresses the problem of being aware of one’s multiple tasks and goals within a dynamic environment and of adapting behavior accordingly.

Turing Machines, Super Turing Machines and DIME Networks:

While a single SPC node lacks self-reflection prohibited by Gödel’s theorems, a network of Turing machines have been successfully used to implement business workflows that observe and manage the external world. This is accomplished by modeling the observed (external to the computing infrastructure) and orchestrating the temporal dynamics of the observed. This has helped us develop complex control systems that can be monitored and controlled with the resiliency of cellular organisms.

However, what is missing is the same resiliency in the infrastructure (or the observer) that implements the control of the observed. Learning fromDamasio’s analysis, in order to introduce consciousness, we must introduce the “self” identity of the observer and the observer’s multiple tasks and goals within a dynamic environment and of adapting behavior accordingly. The “self” specification must include a hierarchy of goals and execution mechanisms to include his concepts of “core” and “extended” selves.

The evolution of computing seems to follow a similar path to cellular organisms in the sense that it emerged as an individual computing element (von Neumann stored program implementation of the Turing machine) and evolved into today’s networks of managed computing elements executing complex workflows that monitor and control external environment.

The Turing machine originally started as a static closed system (Goldin, Wegner, 2008) analogous to a single cell. It was designed for computing algorithms that correspond to mathematical world view. This is the case with Assembler language programming where a CPU is programmed and the Turing machine is implemented using the von Neumann Stored Program Control computing model as shown in figure 2.

Figure 2: A Turing machine with von Neumann Stored Program Control implementation in its simplest form

The Church-Turing thesis stipulates that “Turing machines can compute any effective (partially recursive) functions over naturals (strings). Goldin and Wegner argue that the Church-Turing thesis applies only to effective computations rather than computation by arbitrary physical machines, dynamical systems or humans.

To address this issue, we stipulate that “all computations can be represented as workflows specified by a directed acyclic graph (DAG). Algorithms are a sub set of all computations. An algorithm can be viewed as a workflow of instructions executed by a stored program control (SPC) computing unit (constituting an atomic unit of computation). Then, based on the programming paradigm of one’s choice, one can compose other computing units such as procedures, functions, objects etc., to execute the specified workflow.” This can reconcile the operating system conundrum that states that the operating systems do not terminate as  the Turing machines are required to. As soon as an operating system is introduced, the Turing machine SPC implementation immediately becomes a workflow of computations to implement a process, where each process now behaves as a new Turing machine with SPC implementation. It is as if the operating system is a manager (implementing a management workflow using a group of management Turing machines dedicated for this purpose) controlling a series of other computing Turing machines based on policies set in the operating system. The operating system instructions and the computational flow dependent instructions are mixed to serially execute the process and a sequence of processes. This is analogous to the evolution of multi-cellular organisms where individual cells establish a common management protocol to execute their goals with shared resources. The individual processes may or may not have a common goal but they share the same resources. The operating system communicates with the processes to exert its role using shared memory as shown in Figure 3. While the individual processes do not have fault, configuration, accounting, performance and security management of self, the operating system provides these functions using the signaling abstractions of addressing, alerting, mediation and supervision.

Figure 3: Operating system implements the managed Turing processes.

Since then, multi-threading in a single processor, networked and interactive computing have influenced the computations. In a network, concurrency and influence of one node on another (impact of the environment on the computation) are the new elements that have to be addressed.  The Pi calculus and super Turing models (Eberbach, E., Wegner, P., Goldin, D., 2011) are an attempt to address these aspects. While these attempts are embroiled in controversy, (Cockshott, Michaelson, 2007), what is not in dispute is that a network of computers represents a network of organized Turing machines where each node is a group of Turing machines managed locally. See Figure 4.

Figure 4: A Networked set of Turing machines provide distributed computing services. However this does not provide coordination and management across the two sets of Turing machines.

In such a network, the local operating systems cannot provide Fault, Configuration, accounting, performance and security (FCAPS) management of the system as whole. The disciplines of distributed computing and distributed systems management evolved to address the FCAPS management of the system in an ad-hoc manner without a formal computing model for the system as a whole. This is even more complicated when the system as a whole now acts in unison with a system-wide purpose where one element can influence other elements as pointed out by Louise Barrett (Barrett, 2011).

In this case, the description of the functions performed and the influence of one computation on another has to be encoded at compile time and each computing element does not have the ability to change the behavior at run time. In addition, operating system function is to allocate the resources appropriately to the consumers (processes running applications) and the applications themselves do not have any influence on the resources during run time. For example, if the workload fluctuates, the application has no way of monitoring and controlling the resources.

Figure 5: A network of Turing machines implementing a service workflow that manages the external environment (the observed). The management of the observer is also implemented using the same serial Turing machines where in some nodes the management of the observer and the observed are mixed in serial fashion and some other nodes are exclusively devoted to managing the observer.

If multiple applications are contending for resources, external policies have to be implemented as other Turing machines and the applications themselves are not aware of these external influences. In order to manage distributed set of Turing machines, another set of Turing machines are introduced to provide service management to improve fault, configuration, accounting, performance and security characteristics of the distributed system. See figure 5.

The DIME computing model allows the specification and execution of a recursive composition model where each computing unit at any level specifies and executes the workflow at the lower level. The specification at a higher level eliminates the self-reflection prohibition of Gödel’s theorems on computational units. The parallel implementation of the management workflow and the computational workflow at each level allows the influence of one component in the workflow to influence another component at the lower level. At any level, the computational unit specifies and assures the execution of the lower level workflow thus it becomes the observer observing and controlling the workflow execution at lower level (which is the observed)

This model eliminates the problem of separation of communication between the computing system components in a system and the communication between the computing system and its environment. In current computing models of systems design, treating them as two separate issues has created the current disconnect in the distributed systems theories (Goldin, Wegner, 2007, pp. 22)

Figure 6 shows the new computing model we call distributed Intelligent Managed Element (DIME) network computing model and the resulting computing infrastructure is designed with DIME network architecture.

 

Figure 6: A Distributed Intelligent Managed Element (DIME) with local management of the Turing computing node and signaling channel.  The FCAPS attributes of the Turing node are continuously monitored and controlled based on local policies. In addition the signaling channel allows coordination with global policies.

The DIME network architecture (Mikkilineni 2011) consists of four components:

  1. A DIME node which encapsulates the von Neumann computing element with self-management of FCAPS.
  2. Signaling capability that allows intra-DIME and Inter-DIME communication and control,
  3. An infrastructure that allows implementing distributed service workflows as a set of tasks, arranged or organized in a DAG and executed by a managed network of DIMEs and
  4. An infrastructure that assures DIME network management using the signaling network overlay over the computing workflow

The self-management and task execution (using the DIME component called MICE, the managed intelligent computing element) are performed in parallel using the stored program control computing devices.  The DIME encapsulates the “dispositional know-how.”  Each DIME is programmable to control the MICE and provide continuous supervision of the execution of the programs executed by the MICE. The DIME FCAPS management allows to model and represent dynamic behaviour of each DIME, the state of the MICE and its evolution as a function of time based on both internal and external stimuli. The parallel management architecture allows the observer (a network or subnetworks) that forms a group to monitor and control itself while facilitating the implementation of monitoring and control of the observed in external environment. Parallelism allows dynamic information flow both in the signaling channel and the external I/O channels of the Turing computing nodes.

There are three special features of DNA that contribute to self-resiliency:

  1. Each Turing computing node is controlled by the FCAPS policies set in each DIME. Each read and write are dynamically configurable based on the FCAPS policies.
  2. Each node itself can be a sub-network of DIMES with goals set by the sub-network policies.
  3. The signaling allows dynamic connection management to reconfigure the DIME network thus changing the policies and behaviour.

It is easy to show that the DIME network architecture supports the genetic transactions of replication, repair, recombination and rearrangement.  Figure 7 shows a single node execution of a service in a DIME network.

Figure 7: Single node execution of a DIME

 A single node of a DIME that can execute a workflow by itself or by instantiating a sub-network provides a way to implement a managed DAG  (Directed Acyclic Graph) executing a workflow.  Replication is implemented by executing the same service as shown in figure 8.

DIME Replication

 Figure 8:  DIME Replication

By defining service S2 to execute itself, we replicate S2 DIME.  Note that S2 is a service that can be programmed to terminate instantiating itself further when resources are not available.  In addition, dynamic FCAPS (parallel service monitoring and control) management allows changing the behavior of The ability to execute the control commands in parallel allows dynamic replacement of services during run time.  For example by stopping service S2 and loading and executing service S1, we dynamically change the service during run time.  We can also redirect I/O dynamically during run time. Any DIME can also allow a sub-network instantiation and control as shown in figure 9.  The workflow orchestrator instantiates the worker nodes, monitors heartbeat and performance of workers and implement fault tolerance, recovery, and performance management policies.

Figure 9: Dynamic Service Replication & Reconfiguration

It can also implement accounting and security monitoring and management using the signaling channel.  Redirection of I/O allows dynamic reconfiguration of worker input and output thus providing computational network control.

Figure 10:  Shows DIME Sub-network Implementing Service Composition, Fault & Performance Managements.

Figure 10 shows DIME Sub-network Implementing Service Composition, Fault & Performance Managements. A video link http://youtu.be/Ft_W4yBvrVg provides an animated explanation of the DIME network architecture supporting the genetic transactions of software services implemented using stored program control implementation of the Turing machine.

In summary, the dynamic configuration at DIME node level and the ability to implement at each node, a managed directed acyclic graph using a DIME sub-network provides a powerful paradigm for designing and deploying managed services that are decoupled from the hardware infrastructure management. Figure 11 shows a workflow implementation of monitoring and controlling an external environment (temperature monitoring and fan control to maintain the temperature in a range) using a self-managed DIME network with signaling network overlay.

Figure 11: A workflow implementation using a DIME network. There are two FCAPS management workflows, one managing the observer (computing infrastructure) and the other managing the observed (Thermometer and the Fan)

While the DIME network architecture provides food for thought about Turing, machines, new computing models and the role of the representations of observer and the observed in consciousness, it also has practical utility in developing software exploiting the parallelism and performance of many-core servers (Mikkilineni et. al., 2011). Some of the results demonstrating self-repair, auto-scaling to control the response time of a web server are presented at the Server Design Summit (Mikkilineni, 2011).

Conclusion:

The limitation of Turing Machines as a complete model of computation has been pointed out by (Wegner, Eberbach, 2004). While it was challenged by (Cockshott, Michaelson, 2007), it was rebutted by (Goldin, Wegner, 2008). The main argument for a new computing model was to account for the interactive nature of conventional algorithmic computation and the environment outside the computing element. The Turing model dealing with algorithms is closed and static and does not address the changes affecting the computation from outside while the computation is in progress. In order to account for networked systems in which each change in one element continually influences every other element’s direction of change, more expressive computing model are required. The von Neumann implementation of the Turing machine with its serial processing and mixing of algorithmic computation and interaction using a network of von Neumann computing nodes have given rise to complex management infrastructure that makes it difficult to implement in our IT infrastructure, the architectural resiliency of cellular organisms.

The DIME computing model, by implementing parallel management infrastructure to monitor and control the Turing machine at the atomic level, allows the read and write functions of the conventional Turing machine to be influenced by external interaction. The hierarchical network based (where a node itself can be a sub-network) composition model of DIME network architecture allows the identification of “self” (the observer) at various levels and the representation of the interaction between the observer and the observed.

The beauty of the DIME computing model is that it does not impact the current implementation of the service workflow using von-Neumann SPC nodes (monitoring and control of the observed external systems).  But by introducing parallel control and management of the service workflow, the DIME network architecture provides the required scaling, agility and resilience both at the node level and at the network level (integrating the management and control of self, the observer).  The signaling based network level control of a service workflow that spans across multiple nodes allows the end-to-end connection level quality of service management independent of the hardware infrastructure management systems that do not provide any meaningful visibility or control to the end-to-end service transaction implementation at run time.  The only requirement for the DIME infrastructure provider is to assure that the node OS provides the required services for the service controller to load the Service Regulator and the Service Execution Packages to create and execute the DIME.

The network management of DIME services allows hierarchical scaling using the network composition of sub-networks.  Each DIME with its autonomy on local resources through FCAPS management and its network awareness through signaling can keep its own history to provide negotiated services to other DIMEs thus enabling a collaborative workflow execution.

Each node has a unique identity and supports local behavior and its control using local policies that are programmable using the conventional von Neumann SPC Turing machines. Each sub-network and network allows a group identity (group self) and support group behavior and control.  The resulting network of networks enables system-wide resilient business of managing both the self and the services to monitor and control external behavior. The parallel control network allows dynamic connection management of component functions to create dynamic workflows to accommodate changing environment.

The cellular implementation of the business of managing life may also show us the way to the business of managing our computing infrastructure which has already proven valuable in implementing the business of managing our lives and our environment transcending the body and mind of a single individual. As von Neumann remarked (von Neumann, 1966), “A theorem of Gödel that the next logical step, the description of an object, is one class type higher than the object and is therefore asymptotically longer to describe.” He admitted to twisting the theorem a little while describing the evolution of diversifying computational ecology from simple strings of 0s and 1s (von Neumann, 1987). Perhaps the recursive nature of a network containing sub-networks as nodes along with FCAPS management both at the node and network level, offers the definition of “self-identity” at various levels. While self-reflection at any level is prohibited by Gödel, A higher level “self” provides the required management and control to lower levels. A parallel signaling network, which allows dynamic replication, repair, recombination and reconfiguration, provides a degree of resiliency, efficiency and scaling that are not possible with a network of serial von Neumann implementations of Turing machines only. This may well be a prescription for injecting the property of being aware of one’s multiple tasks and goals within a dynamic environment and of adapting behavior accordingly.

References:

ACM Ubiquity Symposium, (2011) http://ubiquity.acm.org/symposia.cfm

Barrett, L., (2011). Beyond the Brain: How Body and Environment Shape Animal and Human Minds. Princeton, New Jersey: Princeton University Press, p 116, 122

Cockshott, P., Michaelson, G., (2007). Are There New Models of Computation? Reply to Wegner and Eberbach, Computer Journal, vol 50, no, 2, 232-247.

Damasio, A., (1999). The Feeling of What Happens: Body and Emotion in the Making of Consciousness. New York, NY: Harcourt & Company.

Damasio, A. (2010). Self Comes to Mind: Constructing the Conscious Brain. New York: Pantheon Books, p. 25 and p. 35.

Dyson, G. B., (1997). Darwin among the Machines: the evolution of global intelligence. Massachusetts: Helix books, p. 189.

Eberbach, E., Wegner, P., Goldin, D., (2011) Our Thesis: Turing Machines Do Not Model All Computations. (private communication of a unpublished paper)

Goldin, D., Wegner, P., (2008). Refuting the Strong Church-Turing Thesis: the Interactive Nature of Computing, Minds and Machines, 18:1, March, pp.17-38,

Lehrer, J., (2010) How We Decide. Boston, MA: Mariner Books, p. 50

Mikkilineni, R., (2011). Designing a New Class of Distributed Systems. New York,NY: Springer. (http://www.springer.com/computer/information+systems+and+applications/book/978-1-4614-1923-5)

Mikkilineni, R., Morana, G., Zito, D., Di Sano, M., (2011). Service Virtualization using a non-von Neumann Parallel, Distributed & Scalable Computing Model: Fault, Configuration, Accounting, Performance and Security Management of Distributed Transactions, (Preprint)

Mikkilineni, R., (2011). Service Virtualization using a non-von Neumann Computing Model, Server Design Summit (www.serverdesignsummit.com), San Jose, November 29. (A video of the presentation is available at http://www.kawaobjects.com/presentations/ServerDesignSummitVideo.wmv.)

Samad, T., Cofer, T., (2001). Autonomy and Automation: Trends, Technologies, In Gani, R., Jørgensen, S. B., (Ed.) Tools in European Symposium on Computer Aided Process Engineering volume 11, Amsterdam, Netherlands: Elsevier Science B. V., p. 10

Stanier, P., & Moore, G., (2006).  The Relationship Between Genotype and Phenotype: Some Basic Concepts. In Ferretti, P., Copp, A., Tickle, C., & Moore, G., (Ed.), Embryos, Genes and Birth Defects, London: John Wiley, p. 5

Scott, A., (1995). The Controversial New Science of Consciousness: Stairway to the Mind. New York, NY: Copernicus, Springer-Verlag. P.184.

“At the hierarchical level of human conscience it is not possible to report a consensus of the scientific community because there is none. Materialists, functionalists, and dualists are-according to a recent issue of the popular science magazine Omni (October 1993)-engaged in

Slinging mud and hitting low like politicians arguing about tax hikes. Although the epithets are more rarified-here it is “obscuritanist” and “crypto-Cartisian” rather than “liberal” and “right wing”-recent exchanges between neuroscientists and philosophers of mind (and in each group among themselves) feature the same sort of relentless defensiveness and stark opinionated name calling we expect from irate congressmen or trash-talking linebackers.

To the extent that this is a true appraisal of the current status of consciousness, it is unfortunate. Like life, the phenomenon of consciousness is intimately related to several levels of the scientific hierarchy, so the appropriate scientists-cytologists, electrophysiologists, neuroscientists, anesthegiologists, sociologists and ethnologists-should be working together. It is difficult to see how this elusive phenomenon might otherwise be understood.

Davis, P., (1992). The Mind of God: The Scientific Basis for a Rational World. New York, NY: Simon and Schuster.

von Neumann, J., (1966). Theory of Self-Reproducing Automata. Burke, A. W. (Ed.) Chicago, Illinois. University of Illinois Press.

von Neumann, J., (1987). Papers of John von Neumann on Computing and Computing Theory, Hixon Symposium, September 20, 1948, Pasadena, CA, The MIT Press, p454, p457

Wegner, P., Eberbach, E., (2004). New Models of Computation. The Computer Journal, vol 47, No. 1, 4-9.

Is the Software Defined Network (SDN) Another Detour to a Datacenter Dead-end?
August 6, 2012

Introduction

Frustrated by the inability to fiddle with Internet routing in the real world, Stanford computer scientist Nick McKeown and colleagues developed a standard called OpenFlow that essentially opens up the Internet to researchers, allowing them to define data flows using software–a sort of “software-defined networking.” Installing a small piece of OpenFlow firmware (software embedded in hardware) gives engineers access to flow tables, rules that tell switches and routers how to direct network traffic. Yet it protects the proprietary routing instructions that differentiate one company’s hardware from another. SDN is nothing more than the separation of network data traffic processing from the logic and rules controlling the flow, inspection, and modification of that data. Traditional network hardware, i.e. switches and routers, implement these functions in proprietary firmware partitioned respectively into what is known as the data and control planes. While this is a fine research project, as the major vendors start to take this seriously and are attempting to introduce it in the real-world datacenters, one must ask if this will add or reduce complexity in the already complex datacenter where a host of piece meal solutions are offered by mega corporations seeking to continually increase their revenues without an incentive to reduce complexity by eliminating the number of hardware and software components deployed which would cut into their product sales.

Systems theory tells us that as the number of components increase in a system, the cost of complexity could outweigh the benefits unless architectural reorganization provides a way out.  We argue that the management complexity in current IT infrastructure design, based on the serial von Neumann stored program control implementation of the universal Turing machine, is a more fundamental architecture issue related to the lack of resiliency of the computing model than a software design issue. Cockshott et al. (2012) conclude their book “Computation and its limits” with the paragraph “The key property of general-purpose computer is that they are general purpose. We can use them to deterministically model any physical system, of which they are not themselves a part, to an arbitrary degree of accuracy. Their logical limits arise when we try to get them to model a part of the world that includes themselves.” Current generation distributed systems are implemented using a network of Turing machines in which the service and its management are intermixed as shown in figure 1. The resources utilized by the nodes in a network are often controlled by a plethora of management systems which are outside the purview of the service workflow that is utilizing the resources.  Thus the end to end service transaction response is controlled by these management systems which introduce a layer of complexity in coordination and contention resolution making the service much simpler than its management.

Figure 1: Serial von Neumann implementation of Turing Machines

The limitations of the SPC computing architecture were clearly on his mind when von Neumann gave his lecture at the Hixon symposium in 1948 in Pasadena, California (von Neumann, 1987, p. 408). “The basic principle of dealing with malfunctions in nature is to make their effect as unimportant as possible and to apply correctives, if they are necessary at all, at leisure. In our dealings with artificial automata, on the other hand, we require an immediate diagnosis. Therefore, we are trying to arrange the automata in such a manner that errors will become as conspicuous as possible, and intervention and correction follow immediately.” Comparing the computing machines and living organisms, he points out that the computing machines are not as fault tolerant as the living organisms.  He goes on to say “It’s very likely that on the basis of philosophy that every error has to be caught, explained, and corrected, a system of the complexity of the living organism would not run for a millisecond” (von Neumann, 1987,p. 408). It is clear that von Neumann recognized a problem in the way we design computing systems.

“Normally, a literary description of what an automaton is supposed to do is simpler than the complete diagram of the automaton. It is not true a priori that this always will be so. There is a good deal in formal logic which indicates that when an automaton is not very complicated the description of the function of the automaton is simpler than the description of the automaton itself, as long as the automaton is not very complicated, but when you get to high complications, the actual object is much simpler than the literary description.” (von Neumann, 1987,pp. 454-457). He remarked, “It is a theorem of Gödel that the description of an object is one class type higher than the object and is therefore asymptotically infinitely longer to describe.” (von Neumann, 1987,pp. 454-457). The conjecture of von Neumann leads to the fact that “one cannot construct an automaton which will predict the behavior of any arbitrary automaton” (von Neumann, 1987,p. 456). This is so with the Turing machine implemented by the SPC model.

In simpler terms the management complexity is related to the classical Russel Paradox that can be paraphrased as follows: “Who manages the managers?” Gödel’s prohibition of self-reflection in a Turing Machine mandates a hierarchy of Turing machines acting as managers managing other Turing machines implementing the computations described as a sequence of instructions that are compiled into a sequence of 1’s and 0’s. The universal Turing machine (or the general purpose computer) implements these TMs in a synchronous workflow thus prohibiting changes to computations at run-time in any Turing machine while the computation is in progress in that machine (i.e., you cannot change the behavior of that computation (compiled code) till its execution is interrupted).

Current generation server, networking, and storage equipment and their management systems have evolved from server-centric and bandwidth limited network architectures to today’s Cloud computing architecture with virtual servers and broadband networks. During last six decades, many layers of computing abstractions have been introduced to map the execution of complex computational workflows to a sequence of 1s and 0s that eventually get stored in the memory and operated upon by the CPU to achieve the desired result.  These include process definition languages, programming languages, file systems, databases, operating systems etc. While this has helped in automating many business processes, the exponential growth in services in the consumer market also has introduced severe strains on current IT infrastructure. In order to meet the need to rapidly respond to manage the distributed computing resources demanded by changing workloads, business priorities and latency constraints, new layers of resource management are added with the introduction of Hypervisors, virtual machines (VM) and their management. While these layers have made the application or service management more agile, they have introduced a new layer of issues related to their own management. For example, new layers of Virtual machine-level clustering, intrusion detection and performance management, are being introduced in addition to already existing clusters, intrusion detection and performance management systems at the infrastructure, operating systems and distributed resource management layers.

However, this approach is completely unsuited to exploit the new generation many-core servers and high-bandwidth networks now available. The advent of many-core severs with tens and even hundreds of computing cores with high bandwidth communication among them makes the current generation server, networking and storage equipment and their management systems which have evolved from server-centric and bandwidth limited architectures completely unsuited to use in the next generation computing infrastructure efficiently.  It is hard to imagine replicating current TCP/IP-based socket communication, “isolate and fix” diagnostic procedures, and the multiple operating systems (which do not have end-to-end visibility or control of business transactions that span across multiple cores, multiple chips, multiple servers and multiple geographies) inside the next generation many-core servers without addressing their shortcomings.  The many-core servers and processors constitute a network where each node itself is a sub-network with different bandwidths and protocols (socket-based low-bandwidth communication between servers, InfiniBand, or PCI Express bus based communication across processors in the same server and shared memory based low latency communication across the cores inside the processor).

Figure 2 shows the many-core server network supporting multiple bandwidths.

In order to cope with the scaling issues and utilize the hierarchical many-core network of networks effectively, next generation service architecture has to emulate the architectural resiliency of cellular organisms that tolerate faults and implement command and control structures which enable execution of self-configuring, self-monitoring, self-protecting, self-healing and self-optimizing (in short self-*) business processes. This requires new computing models that break the Turing machine barrier to computation by allowing the computer and the computed to be treated in the same model.

Papers Solicited to Address Next Generation Datacenter Infrastructure and Technologies:

The conference on “Convergence of Distributed Clouds, Grids and their Management” sponsored under the Aegis of WETICE 2013 is devoted to addressing next generation computing models which support real-time resource reconfiguration of distributed business workflow execution based on latency constraints, changing workloads and business priorities. It is devoted to addressing the assurance of reliability, availability, performance, account management and security of distributed business process execution with appropriate visibility and control.

The objective of the Conference was first stated in WETICE 2009; “to analyze current trends in Cloud Computing and identify long-term research themes and facilitate collaboration in future research in the field that will ultimately enable global advancements in the field that are not dictated or driven by the prototypical short-term profit driven motives of a particular corporate entity.” We are glad to report that the discussions started in 2009 have directly resulted in an alternative approach to self-managing distributed computing systems totally different from current industry trend showing a way to eliminate the complexity of virtual machines and Hypervisors. If this approach is proven to be theoretically sound (as a paper in WETICE2012 investigated) and extend its usefulness (demonstrated through their feasibility in the form of two proofs of concepts in the last conference) to mission critical environments, the DIME network architecture may yet prove to be an important contribution to computer science.

Following the tradition, the target of the WETICE2013 is to transform current complex, redundant, costly and knowledge intensive IT management into self-configuring, self-monitoring, self-healing and self-optimizing distributed workflow implementations with service management only limited by the speed of light. We identify another emerging area of software defined networks (SDN) as a potential candidate for further investigation without the bias that often surrounds commercial profit motives to see whether the overall complexity of the datacenter will be reduced or the SDNs are yet another layer of complexity.

Papers are solicited to advance the next generation distributed computing and its management infrastructure that leverages the new hardware innovations.  The goals of the conference include (but are not limited to):
  1. Discovering new application scenarios, proposing new operating systems, programming abstractions and tools
  2. Identifying the challenging problem that still need to be solved such as parallel programming, scaling and management of distributed computing elements, and
  3. Reporting results and experiences gained by researchers in building dynamic Grid-based middleware, computing clouds (distributed or otherwise) and workflow management systems.
Submission of papers March 10, 2013
Notification to authors April 1, 2013
Final papers to IEEE-CS April 25, 2013
Paper author’s registration deadline May 10, 2013
 WETICE-2013 Conference June 17-20, 2013

References:

P. Cockshott, L. M. MacKenzie and  G. Michaelson, “Computation and its Limits”, Oxford University Press, Oxford 2012.

J. v.Neumann, Probabilistic logic and the synthesis of reliable organisms from unreliable components, “Automatic studies,” edited by C. E. Shannon, and J. McCarthy, Princeton University Press, 1956, pp. 43-98.

W. Aspray, and A. Burks, “Papers of John von Neumann on Computing and Computer Theory.” Cambridge, MA: MIT Press. 1989.

Cloud Computing, Management Complexity, Self-Organizing Fractal Theory, Non Equilibrium Thermodynamics, DIME networks, and all that Jazz
May 5, 2012

“There are two kinds of creation myths: those where life arises out of the mud, and those where life falls from the sky. In this creation myth, computers arose from the mud and code fell from the sky.”

— George Dyson, “Turing’s Cathedral: The Origins of the Digital Universe”, New York: Random House, 2012.

“The DIME network architecture arose out of the need to manage the ephemeral nature of life in the Digital Universe”

— Rao Mikkilineni (2012)

Abstract:

The explosion of current cloud computing software offerings (both open-sourced and proprietary)  to create public, private and hybrid clouds raises a question. Is it resulting in higher resiliency, efficiency and scaling of service offerings or increasing the complexity by introducing more components in an already crowded datacenter deploying myriad appliances, management frameworks, tools and people, all claiming to help lower total cost of operation? As the reliability, availability, performance, security and efficiency of the total system depends both on the number of components and their configuration, the architecture of a system plays an important role in defining the overall system resiliency, efficiency and scaling. We discuss current cloud computing architecture, the resulting complexity and investigate possible solutions using the self-organizing fractals theory and non-equilibrium thermodynamics. Evolution has taught us that when complexity increases, often, an architectural transformation occurs to lower the overall system entropy. Is a phase transition about to occur in our data centers seeded by the new many-core servers and high bandwidth communications?

Introduction:

According to Holbrook (Holbrook 2003), “Specifically, creativity in all areas seems to follow a sort of dialectic in which some structure (a thesis or configuration) gives way to a departure (an antithesis or deviation) that is followed, in turn, by a reconciliation (a synthesis or integration that becomes the basis for further development of the dialectic). In the case of jazz, the structure would include the melodic contour of a piece, its harmonic pattern, or its meter…. The departure would consist of melodic variations, harmonic substitutions, or rhythmic liberties…. The reconciliation depends on the way that the musical departures or violations of expectations are integrated into an emergent structure that resolves deviation into a new regularity, chaos into a new order, surprise into a new pattern as the performance progresses.” He goes on to explain exquisitely what “all that jazz” means and what it has to do with Dynamic Open Complex Adaptive System or DOCAS.

I borrow the jazz metaphor to understand the current state of affairs in cloud computing. Cloud computing started innocently enough as an attempt to automate systems administration tasks of computing systems to improve the resiliency (availability, reliability, performance and security), efficiency and scaling of services provided by web-hosting data centers. Before the advent of global web e-commerce enabled by broadband networks and ubiquitous access to high-powered computing, the workload fluctuations were not wild-enough to demand very fast response in provisioning to meet them. While enterprise datacenters were not pushed to deal with the wild fluctuations that some web-services companies were, companies such as Amazon, Google, Facebook, Twitter etc., dealing with uncertain (non-deterministic) workload fluctuations took a different approach to improve resiliency and scaling. They took advantage of the increased power in blade servers, high bandwidth networks and virtualization technologies to create virtual machine (VM) based systems administration with multiple VMs in a physical device consolidating workloads that are managed with dynamic resource provisioning. This has become known as cloud computing. Strictly speaking, VM is not essential for automation to improve scaling, auto-failover and live migration of applications and their data; and companies such as Google have chosen their own automation strategies without using VMs. On the other hand, many other enterprises have taken a more conservative approach by not adopting the cloud strategy and avoid the risk of impacting their highly tuned mission critical application availability, performance and security. They are probably correct given the continued occasional outages, security breaches and cost escalation in managing complexity with many public clouds.

Amazon and Google went one step further by offering their flexible infrastructures to developers outside their company to rent the resources with which they could develop, deploy and service their own applications, thus unleashing a new class of developers. Startups could substitute OPEX for CAPEX to obtain the resources required for their new product and services development. Resulting explosion of applications and services has created a new demand for more clouds and more automation of systems administration to extend resiliency and provide a high degree of isolation from multiple tenants sharing resources while resolving the resulting contentions. The result is a complex web of Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS) offerings to meet the needs of developers, service providers and service consumers.  To be sure, these offerings are not independent. On the contrary, each layer influences the other in a complex set of interactions often in non-deterministic way based on workloads, business priorities and latency constraints. Figure 1 shows an example of these relationships.

Figure 1: Complex relationships of information flow between nested layers and information flows between components in each layer. The complexity is only compounded by multi-vendor offerings in each layer (not shown here)

The origin of complexity is easy to understand. While attempting to solve the issue of multi-tenancy and agility, the introduction of the virtual machines gives rise to another complexity of virtual image management and sprawl control. In order to address VM mobility issue, recent efforts to introduce application level mobility using other container constructs such as Gears, Cartridges etc., in the case of Redhat PaaS (or Dynos in the case of Heroku, the salesforce PaaS), introduce yet another layer of management of Gears and Cartridges (or Dynos). Another example is the Eucalyptus Infrastructure as a Service, which goes to great lengths to provide High Availability (HA) of the Infrastructure platform but fails to guarantee HA of applications. It is left to the applications to fend for themselves.  These ad-hoc approaches to automate management have mushroomed the software required, increased the learning curve and made the operation and maintenance even more complex. While all platforms demonstrate drag and drop software with pretty displays that allow developers to easily create new services, there is no guarantee that if something goes wrong, one will be able to debug and find out where the root cause is. Or there is no assurance that when multiple services and applications are deployed on same platform, the feature interactions and shared resource management provided by a plethora of management systems designed independently will cooperate to provide the required reliability, availability, performance and security at the service level. More importantly, when the services cross server, data-center and geographical boundaries, there is no visibility and control of end to end service connections and their FCAPS management. Obviously, the platform vendors are only very eager to provide professional services and additional software to resolve the issues but without end to end service connection visibility and control that spans across multiple modules, systems, geographies and management systems, troubleshooting expenses could outweigh the realized benefits. What we need probably is not more “code” but an intelligent architecture that results in a synthesis of computing services and their management and a decoupling of end to end service connection and service component management from underlying resource (server, network and storage) management.

Self-organizing Fractals and Non-equilibrium Thermodynamics:

Fortunately, the self-organizing fractal theory (SOFT) and non-equilibrium thermodynamics (NET) (Kurakin 2011), provide a way to analyze complex systems and identify solutions. A very good glimpse into the theory can be found in the video (http://www.scivee.tv/node/4994). According to the SOFT-NET theory, the process of self-organization is scale-invariant and proceeds through sequential organizational state transitions, in a manner characteristic of far-from-equilibrium systems, with macrostructure-processes emerging via phase transition and self-organization of microstructure-processes. Once they have emerged as a result of an organizational transition, newborn structure-processes strive to persist and expand, growing in size/number, diversity, complexity, and order, while feeding on pre-existing energy/matter gradients. Economic competition among alternatively organized structure-processes feeding on the same energy/mater gradients leads to the elimination of economically deficient or inferior structure-processes and the improvement, diversification, and specialization of survivors, who are forced to fill and exploit all the available resource niches (the Darwinian phase of self-organization) (Kurakin 2007). Promoted by mutually profitable exchanges of energy/matter, the self-organization of specializing survivors (structure-processes) into larger scale structure-processes transforms (mostly) competing alternatives into (mostly) cooperating complements. As a result, Darwinian competition is transferred onto a larger spatiotemporal scale, where it commences among alternative organizations of self-organized survivors (the organizational phase (Kurakin 2007). Such an economy-driven, scale-invariant process of self-organization leads to the emergence of increasingly long-lived, multi-scale, hierarchical organizations (structures-processes) that expand over increasingly larger scales of space and time, feeding on available energy/matter gradients and eventually destroying them. Yet because energy/matter exists as a non-equilibrium system of interdependent gradients and conjugated fluxes of interconverting energy/matter forms, new gradients and fluxes are created and become dominant as old gradients and fluxes are consumed and destroyed. Such processes are responsible for the continuous birth, death, and transformation of energy/matter forms.

Obviously, cloud computing systems (or for that matter, distributed computing systems in general based on Turing machines) are not living organisms and thus are not susceptible to self-organization. However, if you substitute information to replace energy/matter, there are many similarities between the structure and dynamics of computing systems and living self-organizing systems. The nested computing layers, meta-stable organizational patterns (both macro- and micro- structures) in each layer, and process evolution through inter-layer interaction are the same features that contribute to self-organization. So one can ask what is missing for the cloud computing environments to become self-organizing. The answer lies in two observations:

  1. First one is the Gödel’s prohibition of self-reflection by computing elements that form the fundamental building block in the computing domain, the Turing machine (TM) (Samad and Cofer, 2001).
  2. Second one is the lack of scale invariant macro and micro structure-processes mentioned above for the organization of computing components and their management across various nested layers resulting from current ad-hoc implementation of computing processes using the serial von Neumann implementation of the Turing machine.

I have discussed both these deficiencies elsewhere (Mikkilineni 2011, 2012). The DIME network architecture proposed there attempts to address both these deficiencies.

The DIME Network Architecture:

In its simplest form a DIME is comprised of a policy manager (determining the fault, configuration, accounting, performance, and security aspects often denoted by FCAPS); a computing element called MICE (Managed Intelligent Computing Element); and two communication channels. The FCAPS elements of the DIME provide setup, monitoring, analysis and reconfiguration based on workload variations, system priorities based on policies and latency constraints. They are interconnected and controlled using a signaling channel which overlays a computing channel that provides I/O connections to the MICE (or the computing element) (Mikkilineni 2011). The DIME computing element acts like a Turing oracle machine introduced in his thesis and circumvents Gödel’s halting and un-decidability issues by separating the computing and its management and pushing the management to a higher level. Figure 2 shows the DIME computing model.

Figure 2: The DIME Computing Model. For details on the different implementations of DIME networks (a LAMP stack without VMs and a native Parallax OS) visit http://www.youtube.com/kawaobjects

In addition the introduction of signaling in the DIME network architecture allows a fractal composition scheme of the DIME network to create a recursive distributed computing engine with scale invariant FCAPS management of the computing workflow at node, sub-network and network level. Figure 2 shows the comparison between living organisms with self-organizing fractal attributes and Cloud computing infrastructure organized to exhibit self-management fractal attributes.

Figure 3: Comparison of the nested hierarchical organization of living organisms and DIME network architecture.

While both models exhibit the genetic transactions of replication, repair, recombination and reconfiguration (Stanier and Moore, 2006) (Mikkilineni 2011), there is a fundamental difference between the two. The DIME network architecture is not self-organizing but it is self-managing based on initial policies and constraints defined at the root levels of the hierarchies. These policies can be modified during run time but only with the influence of agents external to the computing element whose behavior is under modification (at the DIME node, sub-network and network level).

At each level, the FCAPS management defines the initial conditions and policy constraints (meta-model if you will, denoting the context and defining the destiny of the ensuing process workflow) that will define the information flows and workflows executed by the DIME network downstream. The resulting metastable configurations are monitored and managed by the managers upstream. This model exhibits the three-step processes that provide self-management in living organisms – establish routine, monitor cues and respond with corrective action based on FCAPS parameters at every level. Figure 4 shows the metastable configuration entropy of the whole system. The FCAPS parameters monitored provide a measure of system entropy shown and the reconfiguration alters the state from higher entropy to lower entropy providing a “measure” of the stable pattern.

Figure 4: System Entropy as a function of time

The SOFT-NET theories provide a path to reexamine the way we design distributed computing systems. Perhaps the living organisms with their self-organizing properties could provide us a way to bring self-management to cloud computing configurations to improve resiliency, efficiency and scaling. The DIME network architecture is a baby-step to implement a recursive distributed computing engine to execute managed workflows that constitute hierarchical and temporal sequences of events executing business workflows.

The DIME network architecture raises some interesting questions about Turing machines and their management. How is it related to the Universal Turing Machine (UTM)? It is important to point out that I do not claim that DIME networks are the answer to Cloud computing vows or that the UTM can or cannot do what a DIME network does. While communicating Turing machines are modeled by a UTM (Penrose 1989), can the managed Turing machine networks also be modeled by the UTM? Is the scale-invariant organizational macro and micro structure-processes discussed in SOFT-NET theory essential for self-organizing systems? What are the differences between living self-organizing systems and self-managing networks? I leave this to the experts. I only point out that the DIME is inspired by the oracle machine discussed by Turing in his thesis and implements the architectural resiliency of cellular organisms in distributed computing infrastructure by introducing parallel management of both the computing elements and networks. While its feasibility has been demonstrated (Mikkilineni, Morana and Seyler, 2012), the DIME network architecture is still in its infancy and presents an opportunity on the eve of Turing’s centenary celebration to investigate its usefulness and theoretical soundness.  Only time will tell if the DIME network architecture is useful in mission critical environments. Figure 5 shows a comparision of Physical server based computing, Virtual Machine based cloud computing and DIME network implementation in Linux server eliminating the Hypervisors and Virtual Machines.

Figure 5: Comparision between conventional, cloud and DIME network computing paradigms. The DIME network Architecture requires no Hypervisors, Virtual Machines, IaaS or PaaS. Linux processes are FCAPS managed and networked using a middleware library without any changes to the Operating System.

The DIME network architecture with its self-management, parallel signaling network overlay and its recursive distributed computing engine model supports all features that current cloud computing provides and more while eliminating the need for Hypervisors, Virtual Machines, IaaS and PaaS. The DNA offers the simplicity by providing FCAPS management of a Linux process through a middle-ware library using standard services of the Linux operating syatem and parallelism available in a multi-core/many-core processor.

Conclusion:

I conclude with one lesson from the past (Mikkilineni and Sarathy, 2009) I take away working in POTS (Plain Old Telephone System), PANS (Pretty Amazing New Services enabled by the Internet), SANs and Clouds. It is that wherever there is networking, switching always trumps other approaches. When services are executed by a network of distributed components, service switching and end-to-end service connection management are the ultimate meta-stable structure-processes and it seems that cellular organisms, telephone networks, and human network eco-systems have figured this out. Signaling and nested FCAPS management structure-processes seem to be the common ingredients. Therefore, I predict that eventually the data centers which are currently computing resource management centers will transform themselves into services switching centers just as in telephony. Perhaps computer scientists should look to telephony, neuroscience and organizational dynamics for answers than engaging in hackathons and coding ad-hoc complex systems to manage distributed computing resources. SOFT-NET theories seem to be pointing to the right direction. The solution may lie in discovering scale invariant micro- and macro structure processes that provide nested FCAPS management and self-managed local and global policy enforcement. Perhaps Holbrook’s “All that Jazz” metaphor is an appropriate metaphor for cloud computing research. Time may be ripe for the reconciliation (the synthesis of the thesis of implementing services and the anti-thesis of services management).

References:

Holbrook, Morris B. 2003. ” Adventures in Complexity: An Essay on Dynamic Open Complex Adaptive Systems, Butterfly Effects, Self-Organizing Order, Coevolution, the Ecological Perspective, Fitness Landscapes, Market Spaces, Emergent Beauty at the Edge of Chaos, and All That Jazz.” Academy of Marketing Science Review [Online] 2003 (6) Available: http://www.amsreview.org/articles/holbrook06-2003.pdf

Kurakin, A., Theoretical Biology and Medical Modelling, 2011, 8:4. http://www.tbiomed.com/content/8/1/4

Kurakin A: The universal principles of self-organization and the unity of Nature and knowledge. 2007 [http://www.alexeikurakin.org/text/thesoft.pdf ].

Mikkilineni, R., Sarathy, V., (2009), “Cloud Computing and the Lessons from the Past,” Enabling Technologies: Infrastructures for Collaborative Enterprises, 2009. WETICE ’09. 18th IEEE International Workshops on , vol., no., pp.57-62, June 29 2009-July 1 2009. doi: 10.1109/WETICE.2009.

Mikkilineni, R., (2011). Designing a New Class of Distributed Systems. New York,NY: Springer. (http://www.springer.com/computer/information+systems+and+applications/book/978-1-4614-1923-5)

Mikkilineni (2012) Turing Machines, Architectural Resilience of Cellular Organisms and DIME Network Architecture (http://www.computingclouds.wordpress.com )

Mikkilineni, R., Morana, G., and Seyler, I., (2012), “Implementing Distributed, Self-managing Computing Services Infrastructure using a Scalable, Parallel and Network-centric Computing Model” Chapter in a Book edited by Villari, M., Brandic, I., & Tusa, F., Achieving Federated and Self-Manageable Cloud Infrastructures: Theory and Practice (pp. 1-374). doi:10.4018/978-1-4666-1631-8

Penrose, R., (1989) “The Emperor’s New Mind: Concerning Computers, Minds, And The Laws of Physics” New York, Oxford University Press pp. 48

Samad, T., Cofer, T., (2001). Autonomy and Automation: Trends, Technologies, In Gani, R., Jørgensen, S. B., (Ed.) Tools in European Symposium on Computer Aided Process Engineering volume 11, Amsterdam, Netherlands: Elsevier Science B. V., p. 10

Stanier, P., Moore, G., (2006) “Embryos, Genes and Birth Defects”, (2nd Edition), Edited by Patrizia Ferretti, Andrew Copp, Cheryll Tickle, and Gudrun Moore, London, John Wiley & Sons

Path to Self-managing Services: A Case for Deploying Managed Intelligent Services Using Dumb Infrastructure in a Stupid Network
February 2, 2012

“WETICE 2012 Convergence of Distributed Clouds, Grids and their Management Conference Track is devoted to transform current labor intensive, software/shelf-ware-heavy, and knowledge-professional-services dependent IT management into self-configuring, self-monitoring, self-protecting, self-healing and self-optimizing distributed workflow implementations with end-to-end service management by facilitating the development of a Unified Theory of Computing.”

“In recent history, the basis of telephone company value has been the sharing of scarce resources — wires, switches, etc. – to create premium-priced services. Over the last few years, glass fibers have gotten clearer, lasers are faster and cheaper, and processors have become many orders of magnitude more capable and available. In other words, the scarcity assumption has disappeared, which poses a challenge to the telcos’ “Intelligent Network” model. A new type of open, flexible communications infrastructure, the “Stupid Network,” is poised to deliver increased user control, more innovation, and greater value.”

                     —–Isenberg, D. S., (1998). “The dawn of the stupid network”. ACM netWorker 2, 1, 24-31.

Much has changed since the late 90’s that drove the Telco’s to essentially abandon their drive for supremacy in intelligent services creation, delivery and assurance business and take the back seat in the information services market to manage the ‘stupid network’ that merely carries the information services.  You have to only look at the demise of major R&D companies such as AT&T Bell Labs, Lucent, Nortel, Alcatel and the rise of a new generation of services platforms from Apple, Amazon, Google, Facebook, Twitter, Oracle and Microsoft to notice the sea change that has occurred in a short span of time. The data center has replaced the central office to become the hub from which myriad voice, video and data services are created, and delivered on a global scale. However the management of these services which determines their resiliency, efficiency and scaling is another matter.

While, the data center value has been the sharing of expensive resources – processor speed, memory, network bandwidth, storage capacity, throughput and IOPs – to create premium-priced services, over the last couple of decades, the complexity of the infrastructure and its management has exploded. It is estimated that up to 70% of the total IT budget now goes to the management of infrastructure rather than to develop new services (www.serverdesignsummit.com). It is important to define what TCO (total cost of ownership) we are talking about here because it is often, used to justify different solutions as the following picture showing three different TCO representations of a data center. Figure 1 shows three different TCO views presented by three different speakers in the Server Design Summit in November 2011.  Each graph, while it is accurate, represents a different view. For example, the first view represents the server infrastructure and its management cost. The second one represents the power infrastructure and its management. The third view shows both the server infrastructure and power management. As you can see the total power and its management, while steadily increasing, is only a small fraction of the total infrastructure management cost.  In addition, these views do not even show the network and storage infrastructure and their management. It is also interesting to see the explosion of management cost shown in figure 3 over the last two decades. Automation has certainly improved the number of servers that can be managed by a single person by orders of magnitude. This is borne by the labor cost in the left picture by Intel which shows it is about 13% of the TCO from server view-point. But this does not tell the whole story.

Figure 1: Three different views of Data center TCO presented in the Server Design Summit conference in November 2011 (http://www.serverdesignsummit.com/English/Conference/Proceedings_Chrono.html). These views do not touch the storage, network and application/service management costs both in terms of software systems and labor.

A more revealing picture can be obtained by using the TCO calculator by one of the Virtualization infrastructure vendors. Figure 2 shows percentage Total Cost of Ownership (TCO) (for a 1500 server data center) over five years by each component with and without virtualization.

Figure 2: Five Year TCO of Virtualization According to a Vendor ROI Calculator. While virtualization reduces the TCO from 35% to 25%, it is almost offset by the software, services and training costs.

While virtualization introduces many benefits such as consolidation, multi-tenancy in a physical server, real-time business continuity and elastic scaling of resources to meet wildly fluctuating workloads, it adds another layer of management systems in addition to current computing, network, storage and application management systems. Figure 3 shows a reduction by 50% of the five-year TCO with virtualization. The Virtual Machine density of about 13 allows a great saving in hardware costs which is somewhat off-set by the new software, training and services costs of virtualization.

Figure 3: TCO over 5 Years with virtualization of 1500 servers using 13 VMs per Server. While the infrastructure and administration costs drop, it is almost offset by the software, services and training costs.

In addition, there is the cost of new complexity in optimizing the 13 or so VMs within each server in order to match the resources (network bandwidth, storage capacity, IOPs and throughput) to application workload characteristics, business priorities and latency constraints. According to a storage consultant, Jon Toigo “Consumers need to drive vendors to deliver what they really need, and not what the vendors want to sell them. They need to break with the old ways of architecting storage infrastructure and of purchasing the wrong gear to store their bits: Deploying a “SAN” populated with lots of stovepipe arrays and fabric switches that deliver less than 15% of optimal efficiency per port is a waste of money that bodes ill for companies in the areas of compliance, continuity, and green IT.”

Resource management based data center operations miss an important feature of services/applications management which is that all services are not created equal. They have different latency and throughput requirements. They have different business priorities and different workload characteristics and fluctuations. What works for the goose does not work for the gander. Figure 4 shows a classification of different services based on their throughput and latency requirements presented by Dell in the server design summit. The applications are characterized by their need for throughput, latency and storage capacity. In order to take advantage of the differing priorities and characteristics of the applications, additional layers of services management are introduced which focus on service specific resource management. Various appliance or software based solutions are added to the already complex resource management suites that address server, network and storage to provide service specific optimization. While this approach is well suited for making recurring revenues for vendors, it is not ideally suited for customers to lower the final TCO when all piece-wise TCO’s are added up. Over a period of time, most of these appliances and software end up as shelf-ware while the venodors tout more new TCO reducing solutions. For example, a well known solution vendor makes more annual revenue from maintenance and upgrades than new products or services that help their cutomers really reduce the TCO.

 Figure 4: Various services/Applications characterized by their throughput and latency requirements. Current resource management based data center does not optimally exploit the resources based on application/service priority, workload variations and latency constraints. It is easy to see the inefficiency in deploying a “one size fits all” infrastructure. It will be more eff icient to tailor “dumb” infrastructure and “Stupid Network” pools specialized to cater to different latency and throughput characteristics and let intelligent services provision themselves with the right resources based on their own business priorities, workload characteristics and latency constraints. This requires the visibility and control of service specification, management and execution available at run time which necessitates a search for new computing models.

In addition to the current complexity and cost of resource management to assure service availability, reliability, performance and security, there is even more fundamental issue that plagues the current distributed systems architecture. A distributed transaction that spans multiple servers, networks and storage devices in multiple geographies uses resources that span across multiple data centers. The fault, configuration, accounting, performance and security (FCAPS) of a distributed transaction behavior requires the end-to-end connection management more like telecommunication service spanning distributed resources. Therefore, focusing on only resource management in a data center without the visibility and control of all resources participating in the transaction will not provide assurance of service availability, reliability, performance and security.

Distributed transactions transcend the current stored program control implementation of the Turing machine which is at the heart of the atomic computing element in current computing infrastructure.  The communication and control are not an integral part of this atomic computing unit in the stored program control implementation of the Turing machine. The distributed transactions require interaction which integrates computing, control and communication to provide the ability to specify and execute highly temporal and hierarchical event flows. According to Goldin and Wegner, Interactive computation is inherently concurrent, where the computation of interacting agents or processes proceeds in parallel. Hoare, Milner and other founders of concurrency theory have long realized that Turing Machines (TM) do not model all of computation (Wegner and Goldin, 2003). However, when their theory of concurrent systems was first developed in the late ’70s, it was premature to openly challenge TMs as a complete model of computation. Their theory positions interaction as orthogonal to computation, rather than a part of it. By separating interaction from computation, the question whether the models for CCS and the Pi-calculus went beyond Turing Machines and algorithms was avoided. The resulting divide between the theory of computation and concurrency theory runs very deep. The theory of computation views computation as a closed-box transformation of inputs to outputs, completely captured by Turing Machines. By contrast, concurrency theory focuses on the communication aspect of computing systems, which is not captured by Turing Machines – referring both to the communication between computing components in a system, and the communication between the computing system and its environment. As a result of this division of labor, there has been little in common between these fields and their communities of researchers. According to Papadimitriou (Papadimitriou, 1995), such a disconnect within the theory community is a sign of a crisis and a need for a Kuhnian paradigm shift in our discipline.”

Kuhnian paradigm shift or not, a new computing model called DIME computing model (discussed in WETICE2010) provides a convergence of these two disciplines by addressing the computing and the communications in a single computing entity that is a managed Turing machine. The DIME network architecture provides a fractal (recursive) composition scheme to create an FCAPS managed network of DIMEs implementing business workflows as DAGs supporting both hierarchical and temporal event flows. The DIME computing model supports only those computations that can be specified as managed DAGs where a management signaling network overlay allows execution of managed computing tasks (executed by a computing unit called MICE) in each Turing machine node that is endowed with self-management using parallel computing threads. The MICE (see the video referenced in this blog for a description of DIME and its use in distributed computing and its management) constitutes the atomic Turing machine that is controlled by the FCAPS manager in a DIME which allows configuring, executing and managing the MICE to load and execute well specified computing workflow and its FCAPS management. The MICE under parallel real-time control of the DIME FCAPS manager aided by a signaling network overlay provides control over start, stop, read and write abstractions of the Turing machine. Two implementations have proven the existence proof for the DIME network architecture.

Figure 5 shows a DIME network implementing Linux, Apache, MySQL and PHP/Perl/Python web services delivery and assurance infrastructure.

Figure 5: The GUI showing the configuration of a LAMP Cloud (Mikkilineni, Morana, Zito, Di Sano, 2012). Each Apache and DNS are DIME aware running in a DIME aware Linux Operating System which, transforms a process into a managed element in the DIME network. A video describes the implementation of auto-failover, auto-scaling and performance management of the DIME aware LAMP cloud

Look Ma! No Hypervisor or VM in My Cloud (See Video)

The prototype implementations demonstrates a side effect of the DIME network architecture, which combines the computing and communication abstractions at an atomic level, – it decouples the services management from the underlying hardware infrastructure management. This makes it possible to implement highly resilient distributed transactions with auto-scaling, self-repair, state-aware migration, and self-protection – in-short, end-to-end transaction FCAPS management – based on business priorities, workload fluctuations and latency constraints.  No Hypervisors or VMs are required. The intelligent management of services workflow with resilient distributed transactions offers a new architecture for the data center infrastructure. For the first time it will be possible to remove embedding service management in the infrastructure management intelligence using myriad expensive appliances and software systems. It will be possible to design new tiers of dumb infrastructure pools (of servers, storage and network devices) with different latency and throughput characteristics and the services will be able to manage themselves based on policies by requesting appropriate resources based on their specifications. They will be able to self-migrate when quality of service levels are not met. The case for dumb infrastructure on a stupid network with intelligent services management puts forth the following advantages:

  1. Separation of concerns: The network, storage and server hardware provides hardware infrastructure management with signaling enabled FCAPS management. They do not encapsulate service management as the current generation equipment does.
  2. Specialization: The hardware is designed to meet specific latency and throughput characteristics to simplify its design through specialization. Different hardware with FCAPS management and signaling will provide plug and play components at run time.
  3. End-to-end service connection FCAPS management using the signaling network overlay allows dynamic service FCAPS management facilitating self-repair, auto-scaling, self-protection, state-aware migration and end to end transaction security assurance.

Figure 4 shows an example design of a possible storage device using simple storage architecture enabled with FCAPS management over a signaling overlay. It can be easily built with commercially off the shelf (COTS) hardware. This design allows separation of the services management from storage device management and eliminates a host of storage software management systems thus simplifying the data center infrastructure.

Figure 5: A gedanken design of autonomic storage and autonomic storage service deployment using the new DIME network architecture. The signaling overlay and FCAPS management are used to provide dynamic service management. Each service can request, using standard Linux OS services during run time, services from the storage device based on business priorities, workload fluctuations and latency constraints.

It is easy to see that the service connection model eliminates the need for clustering and provides new ways to provide transaction resilience with features such as service call forwarding, service call waiting, data broadcast, 800 service call model etc. It is also equally easy to see that with many-core servers, how the DIME Network architecture eliminates the inefficiencies of communication between Linux images within the same container (e.g., TCP/IP) and also how simple SAS storage and Flash storage can replace current generation appliance based storage strategies and their myraid management systems. Looking at the trends, it is easy to see that a paradigm shift soon will be in play to transform the data centers from their current role of being just managed server, networking, and storage hosting centers (whether physical or virtual), to true service switching centers with telecom grade trust. The emphasis will shift from resource switching and resource connection management to services switching and service connection management thus replacing the current efforts to replicate the complexity inside the data center today, also inside the many-core servers. With the resulting decoupling of services management from the infrastructure management, the next generation data centers will perhaps be more like central offices of the old Telcos, switching service connections.

Obviously the new computing model is in its infancy and requires participation from academicians who can validate or reject its theoretical foundation, VCs who can see beyond current approaches and are not satisfied by how many servers can be managed by a single administrator to measure the data center efficiency (as one Silicon Valley VC claimed it as progress in the Server Design Summit) and architects who exploit new paradigms to disrupt the status-quo. The DIME computing model by allowing Linux processes to be converted into a DIME network transcending physical boundaries allows easy migration from current infrastructure to the new one without abandoning legacy applications as the prototype of LAMP cloud demonstrates.

In closing, I like to point out that there have been many calls for a new computing model that combines computing and communication at an atomic computing element level which the Turing machine falls short as discussed above. However, without high bandwidth communication and exploitation of the parallelism that is abundant in the new generation hardware, it is not practically very useful to seriously utilize such new computing models. However, it seems that the hardware advances have outpaced the software advances and perhaps it is about time for computer scientists to seriously take a second look at addressing the software short-fall in dealing with distributed transactions. As the following fable illustrates, it may be futile to look for parallel break-through solutions in a serial boat.

“When Master Foo and his student Nubi journeyed among the sacred sites, it was the Master’s custom in the evenings to offer public instruction to UNIX neophytes of the towns and villages in which they stopped for the night.  On one such occasion, a methodologist was among those who gathered to listen.  “If you do not repeatedly profile your code for hot spots while tuning, you will be like a fisherman who casts his net in an empty lake,” said Master Foo.
“Is it not, then, also true,” said the methodology consultant, “that if you do not continually measure your productivity while managing resources, you will be like a fisherman who casts his net in an empty lake?”
“I once came upon a fisherman who just at that moment let his net fall in the lake on which his boat was floating,” said Master Foo. “He scrabbled around in the bottom of his boat for quite a while looking for it.”  “But,” said the methodologist, “if he had dropped his net in the lake, why was he looking in the boat?”  “Because he could not swim,” replied Master Foo.
Upon hearing this, the methodologist was enlightened”        — Master Foo and the Methodologist
                                                                   (http://www.catb.org/esr/writings/unix-koans/methodology-consultant.html)

If you have transformational research results, or want to make a real difference in computer science research, see Call for Papers at:

www.workshop.kawaobjects.com and http://WETICE.org