Vol 2 No 22
HOT ISSUES
Going Virtual?
Duncan Johnston-Watt, the chief technology officer (CTO), founder and chairman of Enigmatec, has more than 12 years’ experience developing technology for the financial services industry. Recently, Enigmatec made headlines with news of the virtualization project at JPMorgan (DWT, Sept. 26). Over the years, Johnston-Watt has specialized in the development of large-scale systems including the global foreign exchange (FX) and money markets data delivery infrastructure for the French banking giant BNP and the fixed-income analytics infrastructure at UBS. In April 1998, Johnston-Watt joined the Reuters subsidiary Instinet and began the development of a fixed-income brokerage platform that was officially launched in March 2000 (Trading Technology Week, June 19, 1999). By May 2000, Johnston-Watt became managing director, fixed-income technology for Instinet, where he continued to pioneer the use of Java enterprise technologies for financial services. A few months after he left Instinet, Johnston-Watt founded Enigmatec (TTW, May 7, 2001). In recognition of this work, Johnston-Watt was nominated for a Computerworld Smithsonian Award in April 2000. He recently spoke to Eugene Grygo, editor of DWT, about the differences between grid computing and IT infrastructure virtualization.
How do you view grid computing and virtualization?
We see virtualization as an evolution up the value chain from traditional grid computing. In our view, ‘grid’ describes the architecture—an arbitrarily large set of similar producers, each able to provide services to consumers. ‘Virtual’ describes the method of interaction between producers and consumers.
Communication and technology advances enable businesses to launch as ‘virtual enterprises,’ investing primarily in the creation of unique product value, and outsourcing all non-primary business functions, including corporate, sales, marketing, support, and distribution functions. However, virtualization only succeeds if it removes inefficiencies in the value chain to a greater extent than it increases management complexity, or introduces new costs. Many dot-com challengers to existing brick-and-mortar vendors failed precisely when distribution and support costs exceeded facilities and customer acquisition savings.
IT architectures have enjoyed similar virtualization phases. Networks virtualized access to files, storage area networks (SANs), to data, compute grids and to CPUs. Recent virtualization technologies such as VMware virtualize the connection between operating systems and servers, driving out inefficiencies in hardware utilization. Any fixed, inflexible link between components in the data center will eventually be virtualized.
The result? Firms can create ‘virtual services,’ focusing their development energies on solving core business problems, while outsourcing the costly installation, configuration, run-time management, communication, security, and business continuity problems common to all IT applications.
We are at a key transition point in data center virtualization. There are many technology tools available to virtualize nearly every interaction between resources. Each new tool places new education, implementation, and support demands on already over-burdened IT operations teams. IT systems management today relies heavily on human decision-making, customized workflow scripts, collections of vendor-provided management tools, and ad hoc heroics. Management strategies that were developed when applications ran on discrete servers and connections between systems were static are inadequate to handle the demands of highly-distributed, dynamic, round-the-clock virtual environments.
In our view, the only way to drive significant, further costs out of IT management, and realize the benefits of service virtualization, is to reduce the human decision-making bottlenecks from current processes. We must examine the procedures we are convinced can’t be automated, and automate them. Our solutions must be fully distributed (like the services the solutions will manage), flexible enough to handle constant change, reliable enough to survive major component failures, and tied directly to business objectives. This is how we will free IT operations teams from day-to-day fire fighting, allowing them to deliver the next generation of value-added technology solutions.
What are the real differences between virtualization and grid computing? Are they really that different?
No. Grid computing is an implementation of a virtualization strategy. Virtualization breaks fixed, inflexible links between systems and processes, allowing any component to request service from any number of providers of that service. The power of virtualization lies in decoupling consumers from producers, allowing producers—such as a grid of compute engines, or a Web server farm—to scale and be managed independently from the consumers they service.
Can user firms have both—virtualization for one part, grids for another?
Absolutely. This is becoming more and more common. Virtualization can be deployed to increase utilization of an application’s servers, for example, while at the same time a grid can be deployed to improve the performance of a computer-intensive component of the application.
If firms have both, how will they manage the two environments?
Both require new management strategies that can handle real-time change, multiple moving parts, component failure and recovery, and fluctuating demand. Systems management itself must be grid-enabled—fully-distributed, efficient and with decision-making pushed as far out to the endpoints as possible. Virtualized environments demand much higher levels of automated management than systems of the past.
Will firms have to choose between virtualization and grid computing?
No. They work together and will be used to drive the next wave of efficiency from technology resources.
