The Center of the Matter

By Tim Scannell

Kermit the Frog was right.

But despite the complexities of being green--the expenses and the impacts on traditional I.T. strategies and operations--an increasing number of technology vendors and users are taking an eco-friendly approach to data center operations and planning. End users are also starting to swing their purchase votes toward vendors that peddle energy-conserving products that target mobile initiatives as well as centralized mainframes and server farms.

Mobility is a large part of the strategic infrastructure of an enterprise, although most companies have relatively few servers to fully support mobile users and mobile applications, explains Rob Veitch, senior director of business development at Sybase Anywhere. Many companies are looking to add more servers to their I.T. mix but are restricted by rising energy costs, as well as the costs of expanding existing data centers or building new ones.

Sybase, for example, recently announced an initiative to help its customer base use the virtualization techniques and management capabilities that are inherent to its iAnywhere suite of products; the goal is to reduce demands on server resources and thereby reduce the power requirements of host data centers.

"By using the capabilities of these mobile devices you are actually scaling down the power requirements of the enterprise, since most processing is done in batch mode and across multiple systems," says Veitch.

The idea for a more formal approach to mobile energy conservation recently surfaced at a Sybase user group conference where virtualization was a resonating theme, notes the company's business development director. "Sybase is always looking for ways to improve the efficiency of operations and be a good corporate citizen, " Veitch added.

If this sounds a bit self-serving that's all right, since Sybase is not alone in its quest to put a green spin on its gigabyte strategies. Sun Microsystems may very well be the Ed Begley Jr. of computing companies, since it has been pushing energy conservation long before motoring to work in a hybrid car became fashionable. The company has already applied its energy-saving guidelines to its own data centers, replacing 22 outdated servers with 11 newer systems that run slower but can process more tasks. This reportedly cut energy usage by more than 90 percent.

IBM is another long-time proponent of energy conservation in data centers that practices what it preaches. Over the past several years, the company has reduced or consolidated its mainframe and server megacenters from 155 to about seven today.

A decade ago, IBM's data center baggage included more than 15,000 applications. Today, it's closer to about 4,700 applications, says David Anderson, IBM's GreenZ Evangelist.

"The conventional wisdom in the last century was that the distributed model was right because you had a lot of cheap servers," explains . Architectural mindsets changed, however, with advancements in CMOS design, multiple core centralization, cooling techniques and, of course, virtualization. "Now, the trend is to bring everything back to the centralized computer and to do more with intelligent design and microprocessors. The mainframe has evolved to become the hub of enterprise applications," he adds.

This sounds like a complete about-face in terms of computer architectures a decade or two ago, when distributed computing was all the rage--although insists IBM has never "taken its eye off the ball" in terms of mainframe computing. The problem with distributed servers, he says, is that most are incredibly under-utilized, with many operating at as low as 5 percent or less of their potential.

All of this adds up to an incredible waste of energy and resources, as the EPA pointed out in an August 2007 report. An increasing demand for computer resources has led to a significant growth in the number of data center servers and a doubling of the energy required to power and cool these vast server farms. The results are skyrocketing energy costs for businesses and government, an increase in emissions and greenhouse gases and a sizeable strain on the existing power grid.

Processing Pecking Orders

Like other vendors, IBM looked at what was happening in its own backyard when it decided to scale back on its computing megacenters and bring computing home, so to speak, by making use of centralized mainframe solutions. Anderson and others on his green team evaluated some 16,000 servers throughout Big Blue's organization and determined that nearly 9,000 could be eliminated by migrating applications and servers to more powerful mainframe platforms. About 3,900 servers were retired in the first wave, as computing resources shifted to mainframe and more efficient blade server platforms.

Putting things in numerical perspective, at the beginning of 2000 IBM had less than 3.5 million MIPS (millions of instructions per second) of mainframe capacity installed. By the end of the first quarter of 2007, this figure had risen to more than 11 million MIPS, or nearly a fourfold increase, says an IBM spokesperson.

The greening of Big Blue recently led to a more formal plan that kicked off in May 2007 called Project Big Green, a $1 billion per year effort that targets corporate data centers worldwide and involves hundreds of energy architects who are hell-bent on conserving energy in I.T. data centers.

This effort follows a union struck a year earlier with Dell, Hewlett-Packard and Sun Microsystems called the Green Grid Alliance, which was more of an education and awareness-building effort than an ecological call-to-arms.

Positive Reinforcement

Whatever the host structure, some of the most significant impacts of green computing efforts may come when eco-friendly tactics are applied to a mobile workforce, says Sybase's Veitch, who views the millions of enterprise mobile devices as the new model for energy-efficient distributed computing.

"We see a renaissance in distributed computing being driven by a need to provide enterprise solutions at the point of work. You have a lot of data being generated at remote locations," Veitch continued. "You never want to bring that into a single mainframe environment, but should have the appropriate amount of distributed computing power that smooths and filters stuff out before bringing it back to the enterprise." //

Tim Scannell is president of Shoreline Research, a Boston-area consultancy.