IBM Declares War On HP In Blade Server Market

IBM took off the gloves and came out swinging on Thursday when it accused Hewlett-Packard's blade server of being an energy hog. Some industry analysts said the attack was tantamount to declaring war on HP, IBM's chief rival in the blade server market.

In a press conference, IBM said its internal studies show that its BladeCenter system uses up to 30% less energy than HP's BladeSystem. In an industry increasingly strained by astronomical energy costs to power and cool corporate data centers, having that kind of energy savings could be a major win for IBM.

"This is declaring a full-out war and the battlefield is power, cooling, and virtualization," says Dan Olds, a principal with the Gabriel Consulting Group. "Going after them that directly probably means they're pretty sure of their facts, or they're just stupid. And I don't think they're stupid."

HP has become more aggressive in the blade server market and is catching up to IBM, so Big Blue is fighting back, says Jonathan Eunice, a principal IT advisor with Illuminata, an industry analyst firm. "This was unusually aggressive for IBM," he says. "IBM, for a while, pretty much had the blade market to itself while HP went back and retooled. The cClass, HP's most recent blade design, is much stiffer competition for [IBM's] BladeCenter."

In a written response to an InformationWeek inquiry, Eric Krueger, a spokesman for HP said, "We agree with IBM that power and cooling is a big issue for our customers, but, as I'm sure you can understand, without access to IBM's test methodology and results, it's hard to determine if valid, real-world scenarios were used. Knowing what we know about IBM's architecture, it is still an old and outdated design."

Will IBM's test results carry any weight? Many analysts and potential customers are skeptical of studies and tests conducted by vendors, rather than by an objective third party.

But Olds, who has gone over the specifics of the test, says IBM went out and bought an HP blade system and tested it side-by-side with an IBM system. He also notes that IBM used the exact same processors, the same amount of memory, IO devices, disks, and even the same code to generate the workload.

Still, to answer any questions about vendor cheating, Olds says the two systems should be tested by a third-party. "Let's just measure it," he says, adding that IBM probably didn't go after Sun Microsystems or Dell simply because they're not big enough competitors in the blade realm.

The Numbers

According to IBM, its AMD Opteron-based BladeCenter (LS21) within the BladeCenter system uses up to 30% less energy than the comparable HP AMD Opteron-based BladeSystem (BL465c) in the HP cClass system when both systems are idle, and up to 18% less energy when both systems are running at full load.

IBM also reported that the Intel Xeon-based BladeCenter (HS21) within the BladeCenter system uses up to 26% less energy than the comparable HP Intel Xeon-based BladeSystem (BL460c) in the HP cClass system when both systems are idle, and up to 13% less energy when both systems are running at full load. Comparisons are based on systems using the same dual-core Intel Xeon or AMD Opteron microprocessors.

Those could be important numbers for customers battling enormous energy costs and rising temperatures. It's also a market that took in $419 million in the second quarter of this year alone, according to industry analyst firm Gartner.

"As energy bills increase as more and more servers are added, it's becoming more and more onerous over time," says Olds.

In a blade environment, you can put 15,000 or 20,000 watts in one rack, Eunice says. Some financial services companies use between 20,000 and 24,000 watts to run equipment in a single 19-inch equipment rack. "Twenty thousand watts is like a lightning strike," he says. "It's a vast power draw, enough to roast a pig."