Pay Per Sip to Transform IT Economics

In this Issue:

As more and more companies move business applications to the Internet, CIOs are struggling to support applications cost-effectively. They grapple with a patchwork of heterogeneous technologies — both hardware and software — making scaling a major headache. "As company Web sites evolve from information dissemination to transactions and then to business processes, the Internet is indeed becoming a key part to CRM and supply chain management applications," observes Kieran Taylor, director of product management at Akamai Technologies. However, it's increasingly difficult to predict the changing scale of Internet audiences as compared to that of a confined enterprise. "That makes it difficult to know how many servers are needed for how many users when defining capacity for mission-critical applications."

Grid computing holds the potential to "virtualize" computing resources on a global, distributed scale, by giving enterprise executives a means to deploy applications on a "pay-per-sip" model, which transforms IT economics. For some, grid is a "milestone to the Nirvana of using the network as the computer someday," according to Peter ffoulkes, group marketing manager for high-performance and technical computing with Sun Microsystems. He says the advantage to businesses is that grids harness computing resources and data at a more sophisticated level. "That enables companies to harness much more complex information and distributed data for complex transactions and analysis," adds ffoulkes.

Whether it's small servers in clusters or huge symmetrical multprocessing machines, companies choose what's right for their workloads and information supply chains so that they can manage, store, and access data for manipulation and analysis.

Because grid is an operational model — a way of doing computing — it's built on a technology foundation, on top of which business models are built. "The technology foundation essentially is virtualized by having physical resources on top of which Web services technologies are layered," notes ffoulkes. With a set of computing resources, Web services mechanisms can then define business functions independently deployed from underlying infrastructure. "On top is the business model of utility computing where you do a pay-per-use model, so the end result is like a commodity exchange, analogous to the Web where lightweight transactions and analysis take place, except, with grid, it's heavy analysis and transactions that take place."

Grid computing, however, is not easily accomplished. Building reference architectures entails complex business models that must be tested and tuned with the help of partners. "We have customers that have done this multiple times, so they know how to harness computing with a cluster of systems and layer the applications on top," says ffoulkes. However, customers often spend a lot of money and time investing in expertise, which has given rise to customer-ready approaches with plug-in parts. "Then they can have the reference architecture they need and the computer systems that are the best match for their criteria of workload matching, cost, and computers," adds ffoulkes.

Grid computing isn't without drawbacks. One of its dangers is loss of control. Companies must be sure to work with grid computing software that enables them to set limits on how much computing capacity is utilized by setting high and low watermarks on what's to be consumed. "You don't want unpredictability in what you're being billed; you need confidence that there won't be a runaway bill for a given application," warns Taylor.

For that reason, companies should look for software with on-demand pricing models as well as control and visibility of the applications distributed across the grid. Therefore, monitoring and reporting capabilities are paramount.

If done correctly, grid can reduce application server counts significantly while simultaneously improving performance metrics by orders of magnitude.

Susana Schwartz is a New York-based freelance writer specializing in emerging technologies and their impact on IT infrastructure.