Street Makes Forays Into Utility Computing

By Maria Trombly | Securities Industry News | June 16, 2003

The promise of utility computing is that computer power could be delivered as easily as electricity. You get exactly as much as you need, you only pay for what you use, and, except for hot summer days in California, there's always more available if you need it.

The problem with that analogy is that one watt is much the same as another. Power companies don't run benchmarks to show that their electricity is more efficient or faster than that of their competitors. In addition, there's a standard for power plugs, said David Dibble, EVP of technology services at San Francisco-based Charles Schwab. "Electricity depends on a highly standard interface," he said.

Computers, on the other hand, are very differentiated, with different hardware platforms, different operating systems, and, of course, different applications requiring different ways to plug into each other. Furthermore, there's no actual way to virtually deliver computing capacity. There's no way to instantly bring in extra power to crunch end-of-year numbers the same way you bring in extra power for the Christmas lights.

What you can do is bring in a technician to install a few more processors or another server, or send the application somewhere else to run.
There's also a third approach: shifting the load between different machines you already own. All three of these methods are referred to as utility computing.

"When you talk about utility computing or on-demand computing, these are still nebulous terms that can mean different things in different contexts," said Jason Bloomberg, an analyst at Waltham, Mass.-based ZapThink Llc.

Wall Street firms are looking at all the possible approaches, however, because utility computing can save money and improve flexibility, he said. "The cost savings come fundamentally because when you go out and buy computers, you have to buy enough computers to handle peak demand," he said. "That's usually the most important time for your systems to work. That means that, most of the time, they're sitting mostly idle. That's money being wasted. Utility computing is a way to deal with those peak demands."

On the flexibility side, the ability to allocate resources to applications as needed makes it easier for companies to handle changing regulatory requirements or market conditions, he added. "All the financial firms are now struggling to comply with Sarbanes-Oxley," he said. "Is this going to be the last regulatory action that's going to shake things up? No. There's always something that's going to hit the industry and they'll have to respond."

Because of the potential cost savings and increased flexibility, Charles Schwab, for one, is firmly committed to the future of utility computing, according to Dibble.

"We very much believe in the promise of dynamic provisioning and autonomic systems," he said. "What the challenging business climate has done for us is to put an incredible focus on cost-effectiveness. During the '90s, we were still concerned about how to spend money, but with the market growing the way it was then, you just had to keep up. Now, there's been an immense focus here at Schwab and at other places, and in the vendor community, on what you can do to build the most efficient infrastructure possible."

IBM is currently the vendor that's most ahead of the pack when it comes to delivering on the promise of utility computing. Sun Microsystems, Hewlett-Packard and Computer Associates, as well as a number of smaller firms are working on the management problems that utility computing poses. And a related effort is under way to virtualize storage by making it easily accessible through a network.

As part of these efforts, HP, for example, has created a new unit-a "computon"-roughly equivalent to the way the watt is used to measure electricity. The computon will be part of a metric to determine how much to charge outsourced utility computing customers. Unlike the watt, however, the computon isn't based on an actual physics concept and so will have to include a number of variables such as performance.

Despite all the talk of outsourced utility computing, however, the most common application of utility computing today is within a corporate firewall, redistributing resources on the fly to applications that need them most.

"Wall Street's IT infrastructures over the last five or 10 years have grown dramatically," said Dan Powers, VP of grid computing strategy at IBM. "And huge amounts of complexity have been built into that environment, as well as huge inefficiencies. If you look at your Windows desktops or servers, the CPUs are used maybe 5 percent over a 24-hour period. Unix servers are 15-percent to 20-percent-used."

Utility computing can be used to draw on some of these idle resources to run a number of applications, including risk management, portfolio optimization and even transactions, said Powers.

A good utility computing system can also reduce management costs, Powers added. Systems administration can take up as much as half of the total IT budget, he said. "It's becoming hard for people to be responsive to their business when all they're doing is spending their time keeping the systems up and running."

One of the ways that utility computing is approached within the enterprise is grid computing, which lets a company put together a large number of small machines into one large virtual supercomputer. Also called distributed computing, this approach is best known for letting millions of individual Internet users help in the Search for Extraterrestrial Intelligence (SETI) by running a free program that downloads and analyzes radio telescope data.
More down to earth, Wall Street firms use grids to run compute-intensive applications, such as portfolio analysis. Distributed computing also lets a company scrounge spare processing time from computers that would otherwise sit idle.

After all, computer time-like airplane seats-is a perishable resource. A computer depreciates on a daily basis, whether or not it's actually being used.
Specialized vendors like New York City-based DataSynapse help Wall Street customers run applications on grid systems. Firms like Morgan Stanley, CitiGroup, Prudential, Deutsche Bank and First Union Corp. have already put the technology into practice.

DataSynapse recently partnered with Sun, which, in addition to its own Grid Engine product will also be offering the DataSynapse grid products to its customers.

At Charles Schwab, for example, grid computing allows an adviser to run a portfolio simulation in 15 seconds-instead of the four minutes it took previously.

"It's a big difference of sitting on the phone for four minutes twiddling your thumbs, and 15 seconds," said IBM's Powers.

According to Sch-wab's Dibble, IBM first came to Schwab with the idea to start moving toward utility computing about a year ago, and the grid system is expected to go into production by year-end.

"We're really happy with the results," Dibble said. "It's been wildly successful. The grid lets us deploy very computer-intensive applications on existing machines that were actually put in place for different things. The main purpose of the machines was not interfered with at all."

The most interesting thing that's now happening in utility computing is the advent of service-oriented architecture, exemplified in Web services.
A Web service is an application that, on the outside, speaks the universal language of Soap (Small Object Access Protocol) and XML. Web services aren't identical to one another the way all watts of electricity are the same, but they're the closest we've come so far. Many Web services are also platform-independent. With the exception of Microsoft, most vendors with Web services development platforms use J2EE, a language that can run on any machine-Unix, Linux, Windows and everything else out there.

This makes it possible to take a Web service-such as looking up a stock quote-and move it to additional servers, or even offsite to an outsourcer. The fact that Web services communicate with one another in standard ways makes it possible to do this without breaking up its connections to other applications.

Cap Gemini Ernst & Young Llc calls this "micro-utilities."

"With at least 20 percent of our clients in financial services, we are working with them to explore, architect and determine if they should go forward with the idea of a micro-utility," said James Greene, the company's global head of financial services consulting.

This trend, also called business process outsourcing, is expected to grow as Web services standards develop further and become more robust. "Service-oriented architecture is the organizing principle for this computing-on-demand world," said Zapthink's Bloomberg.

According to Greene, the first step for a Wall Street company is to implement service-oriented architecture internally. Then, when all the connections use Web services communication standards, any individual Web services can be pulled out and outsourced.

According to Greene, Cap Gemini is already working with Wall Street clients on moving Web services to third-party outsourcers. Cap Gemini will also serve as the owner and operator of some "micro-utilities" for its Wall Street customers, he added.

Being able to pick the best possible location to run a Web service-either internally or from an outsourcer-allows firms to not be locked into any one particular vendor, so they can look for the best price, the best performance or some other combination of requirements. At least, that's the plan.

"You have to be careful about separating the practical use from the hype," warned analyst Bob Parker at Boston-based AMR Research. "It's extremely interesting but there's a barrier in that the applications themselves need to be designed and built so that you can provision them, and the application providers aren't there yet. But, all that being said, there is a message here that resonates not just on Wall Street but for manufacturing and retail as well, around this notion of the utility-not just being able to ramp up capacity when I need it but also combining it with outsourcing that's based on variable costs based on how much I use it."

There's no question that this is the way Schwab wants to go, Dibble said.

"Service-oriented architecture allows you the flexibility to source from many different locations," he said. The firm is beginning to implement this style of application development internally. "Schwab absolutely has service-oriented architecture," he said.

However, it's still too early to actually put it to work for external outsourcing, he added. "There's still a very pliable environment," he said. "The standards are still being developed."

And while Web services are an important step toward utility computing, he added, they don't eliminate all the variety in the way applications interact.

"Web services alone will not get us there," he said. "We need a consistent application architecture; that's probably the single most important consideration."

But he added that he expects to see this kind of outsourcing happening pretty soon. "Certainly within the next year, the possibilities for that are better than ever."

DEFINITIONS

Utility computing: A technology that allows computing power to be delivered at a variable rate, either by drawing on excess capacity elsewhere in the enterprise or by drawing on capacity available from an outsourcing firm.

Grid computing: Also known as distributed computing, this technology allows a number of small machines to be brought together into one virtual supercomputer. A grid can either use dedicated servers or draw on unused capacity available on existing machines. Utility computing often uses grids as an inexpensive source of computing power.

Service-oriented architecture: A way of organizing applications so that the business logic is broken up into small, reusable chunks, known as services, which interact with each other using standard messaging formats.

Business-process outsourcing: A process in which individual services or groups of services are lifted up out of a firm's service-oriented architectural framework and moved to an outsourcer without interruption to the company's business logic.