Hardware Today: Server Room Crystal Ball

by Drew Robb

By any measure, 2004 wasn't a year of massive change in server rooms. We check in with the analysts to see what 2005 has in store for blades, dual-core processors, server spending, and more.

By any measure, 2004 won't go down as a landmark year in the evolution of the server marketplace. Sure, everything got a little better — more powerful processors, more RAM, improved architectures, and enterprises that spent more money, but vendors largely served up incremental gain, rather than the proverbial better mousetrap.

"2004 saw more maturity in blade technology — specifically support for more I/O options from IBM and HP — plus Dell re-entered the market with a new blade product," said Gartner analyst John Enck. "We also saw more attention focused on rising power and cooling demands of x86 servers of all form factors."

Therefore, rather than looking back at the fairly ho-hum 2004, we're opting to look ahead and see where these emerging trends may be taking us in the year to come.

Mandatory Server Upgrades

The coming year could see a bonanza in server replacement, as well as consolidation of multiple small units into one or two larger boxes.

Many companies invested in new infrastructure and servers in the run up to Y2K. Then, hot on the heels of the "IT apocalypse," came the dot-com bust, and hardware spending has been sparse for the past several years. During 2004, budgets began to loosen up a little, and that trend is likely to continue in 2005.

"Many companies are now looking at server upgrades and server consolidation," said Clive Longbottom, a business process analyst at the U.K.-based IT consulting firm, Quocirca.

Ideally, hardware should be replaced every few years. With Wintel boxes in particular, three years can render the equipment near-obsolete. Many organizations, though, are still clinging to stuttering (and soon to be unsupported) NT boxes. The coming year could see a bonanza in server replacement, as well as consolidation of multiple small units into one or two larger boxes.

Blades in the Shade?

Some regard 2004, as the end of the server blade honeymoon. The novelty value of the vision has worn off, and serious real-world issues, such as heating and cooling, have surfaced. Toward the end of the year, though, blades gathered more momentum, indicating they will be a major force in 2005 and beyond. Most recently, RLX, the vendor that coined the product name, announced it will no longer sell the servers themselves and will instead focus on the management software that drives the blades.

"The blade market has been relatively dormant and vendors haven't marketed their products particularly well," said Longbottom. "As we move towards more of a virtualized and service oriented world, the blade makes much more sense, and I think we will see an uptick in this area next year."

Gartner's Enck agrees.

"The coming year will see continued evolution and vendor pushing of blade technology," said Enck. "Look for notable improvements in the software stack that manages blades."

Dell, for example, recently re-entered the blade market. The PowerEdge 1855 has increased density and a lower total cost of ownership than traditional 1U rack servers. Dell's architecture supports up to 10 servers in a 7U chassis. According to Antonio Julio, from Dell's PowerEdge server marketing group, the blade server consumes 13 percent less power and generates less heat than Dell's 1U servers, which helps reduce operation costs.

"We feel that blades will continue to gain traction this year and become even more widely deployed," said Julio.

>> Dual Core, Convergence, the End of the OS?

Dual Core Dueling

Several analysts predict that the introduction of multicore processors will shake up the market. Enck expects this to lead to a new price/performance point in the x86 server market in particular.

Others, though, foresee problems arising from dual-core Intel at the licensing level.

"The software vendors refuse to get their act together and will insist on charging as if these dual cores are two processors," Enck said. "With the sum of the parts only being around 1.6 to 1.8 rather than 2, the buyers will figures out they are again being fleeced, and that will lead to different software models being tried out by the software guys."

Server Convergence

A few years ago, Web servers and application servers were very different animals. That has been shifting slowly, and in 2005 it will change even more.

"We are definitely seeing Web servers, application servers, and integration servers [combine] into one product," said Dennis Byron, an analyst at IDC. "This won't be completely done by 2005, as the server battleship doesn't complete a maneuver in one year, but it is gathering momentum."

He sees this trend spilling into the open source market. As more vendors introduce Linux products, they will have to find ways to differentiate their wares. An obvious way to do this is by adding more tools. A transaction monitor built into an application server, for example, might enable some organizations to cut costs by getting off CICS or Tuxedo.

The operating system may become something of an anachronism. After all, the application needs only a JVM to run, and a JVM on silicon removes many layers of code and possible frailty from the overall equation.

Virtualization is also enabling this to happen. Here, vendors like IBM and BMC are working intently on technologies that drill down to the hardware itself. This results in being able to rapidly provision and de-provision hardware with the correct operating system and application server stack.

"IBM and BMC say they are close to bringing this down to a couple of minutes," said Longbottom. "If they achieve that, the virtualization of the Intel-based server environment will follow rapidly."

The End of Operating System Differentiation

Will 2005 finally be the year that Linux makes a serious dent in the current Windopoly? Most analysts expect Linux to continue to gain ground in the server space. But some anticipate some more problems for open source as it strives to capture market share from Windows and Unix.

"Some Linux deployments haven't gone well, and that may result in some pull back from the platform," said Rob Enderle, an analyst with Enderle Group. "The next two years will be a knock-down, drag-out fight between Windows and Linux and will set the stage for a very interesting 2007, when Microsoft's next-generation server platform is expected to debut."

By then, though, might the operating system be playing a less important role in IT overall?

Longbottom see this as a distinct possibility based on the fact that today's applications are being written with an application server platform in mind. As a result, the operating system may become something of an anachronism. After all, the application needs only a JVM to run, and a JVM on silicon removes many layers of code and possible frailty from the overall equation.

"There is a possibility that we may see the first signs of the death of the operating system in 2005," concludes Longbottom.

This article was originally published on Tuesday Dec 28th 2004
Mobile Site | Full Site