SHORT ATTENTION SPAN SUMMARY
Telecommunication systems have evolved over a century to develop a reputation for service reliability often referred to as “Telecom Grade Trust”. The Internet embraced many of the concepts from Telecommunications Network to establish the same level of trust and provide data and voice convergence. The Internet has since allowed many services to be delivered that facilitate communication, collaboration and commerce at the speed of light. However, these services are developed and delivered from the Information Technology infrastructure that has different characteristics from the Internet and the Telecommunications infrastructure. This explains why many of the services such as e-mail from Google, or e-mail delivered through Blackberry handheld device do not offer a similar level of Telecom-grade trust.
In the past, the mainframe systems did provide a high degree of reliability, security and performance, but had other problems with agility, linear scaling and high cost. This led to the current IT systems administration paradigm that evolved over the last forty years with an open-systems approach. It is based on the assumption that the time-scales of systems administration are slow enough to tolerate human latency. However, the evolution of communication, collaboration and commerce at the speed of light is demanding a systems administration process that allocates the right resources to the right application at the speed of light based on business priorities and changing workloads.
The dissonance between services transcending space and time boundaries and their management requires a re-engineering of current IT infrastructure to eliminate the latency. Just automating current processes only adds to the complexity and both the system and software vendors have to reexamine the administration paradigm to leverage the advances in multi-core, multi-CPU hardware and virtualization technologies.
The 2nd IEEE International workshop on collaboration and cloud computing is challenging the academicians and industry researchers to develop new architectural simplification of the solutions.
SOME THOUGHTS FOR REFLECTION:
As a colleague of mine from AT&T Bell Labs used to say, “Software is human knowledge encapsulated in executable form as opposed to manuals where it is encapsulated in non-executable form requiring a human agent to execute.” During the past fifty years, as more and more human knowledge from many domains is translated into executable form using workflow automation, the way we communicate, collaborate and conduct commerce has transcended space and time boundaries only limited by the speed of light. While Information Technologies (IT) have facilitated this transformation, many have started to question their effectiveness. Instead of repeating the same arguments about Return On Investment and Total Cost of Ownership conundrum, lack of agility or how 70% of IT investment goes into its own maintenance thus robbing resources from new solutions development, I would like to follow Arnold Kling and Nick Schulz [1] in identifying the “intangible assets” and ”the hidden liabilities” in today’s IT infrastructure. It suggests some new directions for the systems and software evolution and the hope is that the participants of the Second International Workshop on Collaboration and Cloud Computing will discuss some new ideas.
According to King and Schultz, the “intangible assets” are knowledge bases. This category includes formal scientific findings and technological innovations. Multi-CPU, multi-core Servers, virtualization technologies, mobile technologies, green technologies and software innovation fall into this category. It also includes less formal learning from experience, such as the know-how and workflows that improve productivity (automated or otherwise). “Hidden liabilities”, on the other hand, are institutional and cultural impediments to innovation and productivity. These range from structure and conduct of government to the attitudes and customs of ordinary citizens.” We can include current vendor reluctance to change status quo, short-term economic pressures from the Wall Street casino, embedded legacy processes and plain inertia that discourages change to the list. Economics 2.0, as they call it, emphasizes the role of innovation and entrepreneurship to overcome the market failure to change from status quo and incremental innovation to a paradigm shift and order-of-magnitude improvements.
The authors claim that “conventional economics is focused on how we can allocate resources efficiently. In this view, the story is that markets facilitate trade and thereby foster efficiency but do little else. With painstaking graphs and numerical examples, the professor shows that both sides of trade benefit from the exchange, whether trade takes place within a national border or across it. These calculations explain why it is better to outsource your ironing to a laundry than to do it yourself. Economics 1.0 explains that trade is based on comparative advantage. Economics 2.0 says, yes, it is more efficient to send your shirts to a laundry than to iron them yourself. But have you heard of permanent press? Thanks to technical progress, many shirts today do not need to be ironed at all. Perhaps in another decade, or two, they will not need to be washed. Given the likely progress of nanotechnology, there is a good chance that the shirts manufactured in 2020, will be “permanent clean.”
Current open systems approaches that profess choice and diversity have also introduced complexity. Any systems administrator will immediately acknowledge the complexity of diagnosing troubles and the amount of time spent on dealing with the idiosyncrasies of different hardware and software systems. Despite vendor claims of one-click installations, configuring new hardware and software takes trial and error, visits to many web-sites for answers to similar problems, thick volumes of reading material and multiple downloads of patches from different vendors. Recently, I spent three weeks to install and make Windows server 20008 R2 work on a Dell PowerEdge 2970 and another four weeks to bring up Ubuntu on the same machine as a virtual guest OS inspite of excellent support from Dell and very good design of both Windows and Ubuntu systems. The productivity loss in the universe is directly proportional to the number of systems, even granting the new automation assistance from various sources.
The current approach to throw software and systems out and offer services for recurring revenue to assist customers and make them operational is not a scalable approach as more consumers become computer sytems dependent, and neither is asking each consumer to become a geek to use the technology. Unless we find ways to create systems that are self-configuring, self-monitoring, and self-healing, the benefits of intangible assets in current IT industry will be outrun by the costs of hidden liabilities.
Both POTS (Plain Old Telephone System) and PANS (pretty Amazing New Services) infrastructures have evolved to embrace many of the self-configuring, monitoring and healing attributes through Intelligent Network (IN) architecture and distributed systems management. It is time to learn from the past and reexamine current IT infrastructure evolution. With the advent of Virtualization, multi-CPU and multi-core servers and our current understanding of distributed computing and operating systems, it is possible to reexamine the systems administration paradigm and eliminate various steps we take today (albeit automated) and re-engineer the IT infrastructure to eliminate unnecessary human latency.
Evolution from POTS, PANS and SANs to Intelligent Collaborating Cloud Network (ICCN)
Figure 1 shows the evolution of current communication and computing infrastructure. The Internet combined with virtualization technology allows the software systems to transcend physical, geographical and time boundaries to enable human communication, collaboration and commerce at the speed of light. However, tying the new technologies to legacy evolution only slows down the growth. What we need is a bold and new approach that leverages the new technologies to progress rapidly while wrapping the legacy to integrate smoothly with future advances. We have done this before with COBOL programming, digital switching and wire-line to wireless transitions.
Now imagine you are a vending machine and the supply of the soft drink you are dispensing is running out. You would like to call the local supplier who can replenish your supply just in time. Imagine that you are being vandalized and you would like to call for help. Imagine now, you are a vendor providing a group of vending machines in a mall. Wouldn’t it be economical to create a local gateway that keeps an eye on the safety and supplies of the group? By the way, you can also remotely monitor, manage and optimize your small vending cloud.
You can imagine similar scenarios with home computing, entertainment, education, healthcare, security monitoring, and remote control systems. When a new device or a service is added or removed, you would like it to be added or removed transparently to the management systems that provides local Fault, Configuration, Accounting, Performance and security (FCAPS) optimization and collaborate with remote distributed entities to extend FCAPS management globally.
On a larger scale, you can imagine local power grid systems being FCAPS managed locally and being connected with other distributed grid systems to provide global management and optimization strategies. These scenarios are possible because of the deployment of intelligence in distributed computing elements with networking and storage, which can communicate with other elements on demand.
With collaborating computing clouds, it will be possible to provide same features that Wall Mart and other large organizations use to small businesses without breaking their bank accounts. Management of Power Grids will be as easy as current telecommunications management systems that assure voice and data with “Telecom Grade Trust”
The above examples illustrate the more general distributed computing services network model where managed workflows are implemented using clusters of computing networks organized in various topologies that are optimal to support the overall end-to-end objectives taking into account service latency tolerance. Four major technology advances have made this architecture efficient and effective:
- Proliferation of powerful Commercially Off The Shelf (COTS) computing processors, network connectivity and storage at affordable cost embedded in a variety of edge devices to monitor and control local environments
- Higher bandwidth networks and ubiquitous wire-line and wireless connectivity allowing workflow implementations with desired latency tolerance
- Virtualization technology that no longer requires application software (used to implement services executing the components of a local or global workflow) to be tied to a particular physical infrastructure in a particular location and
- Distributed computing software advances that provide latency tolerance, high availability, end to end security, performance optimization and improved resource utilization
The combination of virtualization and the broadband Internet have taken the Information Technology (IT) hosted managed services to a next level of evolution, where the software applications have become independent of the hardware infrastructure and can be migrated at will. In addition, the service creation, which once was in the purview of IT software development groups exclusively, has been democratized and anyone with the access to the network can now start developing new services on virtual servers and deploy them in the market place to meet wildly varying workload conditions. This is very similar to the Intelligent Network deployed to deliver voice and data services in the past with service creation, delivery and assurance platforms to connect billion of users. The following figure shows the old IN model modified to show the new trend.
This new architecture fundamentally alters the way in which current IT infrastructure is deployed by virtualizing all the functions that are hardware or appliance oriented such as load balancers, firewalls, and even routing functions. In the limiting case, it is easy to imagine that all physical computers, network elements and storage systems becoming completely commoditized just like washers and dryers where virtual services are being deployed through virtual servers with specific functionality and appropriate latency management. This allows implementing business workflows using managed distributed computing elements with unparalleled agility.
The distributed computing model that allows individuals to form groups, set goals and implement workflows leveraging specialization and separation of concerns has been evolving over centuries to create distributed human networks and allowed groups, organizations and even nations to scale with a purpose. The main purpose of the distributed computing networks is to optimize the distributed resource utilization to accomplish the goal. In its essence, distributed computing enables execution of workflows by a network of distributed agents to accomplish the overall goal of the group. In most cases, the workflow tasks involve:
-
Monitoring local environments by gathering information
-
Pre-process, filter and prepare the information to be correlated or used by other agents
-
Make decision to provide local control and optimization based on global policies
-
Correlate and analyze the information gathered by different agents in light of group’s objectives
-
Make relevant decisions based on the analysis
-
Request actions based on the decisions from different agents and
-
Execute various actions based on the requests to control or manage local environments
While virtualiation technology commoditizes the computing hardware, by making the service execution independent of underlying hardware, by allowing movement at will, the distributed computing network implementation provides a new opportunity to the chip and system vendors to support features that allow the implementation of dynamic local FCAPS management in hardware, hypervisors and operating systems thus providing efficiency and global interoperability. This will allow creating local services with dynamic FCAPS management and a managed network of distributed computing services. By combining virtualization and the distributed computing network model, we can create clusters of computing networks interconnected to implement managed workflow implementations with latency tolerance to meet the business objectives. This is equivalent to implementing distributed computing clouds, each cloud offering the services of a mini-data center where local workflows are implemented and managed through coordination and collaboration with other mini-clouds.
The chip and computer vendors can play a substantial role in simplifying the implementation of these services by incorporating specific service enabling features that support local FCAPS management and interoperability with other clouds. This has been done before in telecommunication networks and IP service networks. For example when the routers were first introduced routing tables had to be manually entered. Every time the router was switched off, the tables had to be reentered to start the service. FCAPS management was added as an afterthought to reduce the tediousness of operation and management. In telephony, dynamic provisioning altered the service economics substantially. Similarly, if the cloud appliances (with computing, networking and storage services) are not designed with operation and management ease at the chip level, and appropriate operating systems are not designed to address FCAPS management of virtual servers, distributed computing networks cannot scale to implement distributed managed workflows that address latency issues. Will this provide the chip vendors and hardware vendors a completely new opportunity? Are there solutions that go beyond automating existing processes that evolved over last four decades incrementally adding enormous complexity? Is network-centric resource management different from server-centric management we are used to today?
One clear lesson we can take away from POTS, PANS and SAN implementations is that when resources are distributed and shared, they need to be managed to compose them and provide servics to their consumers with appropriate end-to-end service level assurance. Service consumers and resources require mediation to resolve contention for shared resources and resolve conflicts based on global business priorities. Services that enable communication, collaboration and commerce at the speed of light demand dynamic services management at the speed of light.
A successful implementation of IN required a smooth integration of hardware and software features that provide dynamic resource management. Virtualization offers a new opportunity to provide dynamic IT resource management. The 2nd International IEEE Workshop is intended to continue the discussion of some of the innovative approaches started in the last workshop [2] by bringing both the academic researchers and industry R&D leaders together. Now let us go back to the first picture you saw. If you see a way to go from the present to the future in the picture, you are cordially invited to contribute a paper to this workshop and try to establish leadership in creating the next generation cloud computing. You may change the game and prove the postulates of Economics 2.0.
Refereces Cited:
[1] Arnold Kling and Nick Schulz, “From Poverty to Prosperity: Intangible Assets, Hidden Liabilities and the Lasting Triumph Over Scarcity”, Encounter Books, New York, London, 2009
[2] P. Goyal, “The Virtual Business Services Fabric: an integrated abstraction of Services and Computing Infrastructure,” in Proceedings of WETICE 2009: 18th IEEE International Workshops on Enabling Technologies: Infrastructures for Collaborative Enterprises, 33-38 (2009).
