With no one else to beat up in the blade server market, HP and IBM have decided to make a sport of bashing each other.
Last week, HP highlighted a report from Sine Nomine Associates – your guess is as good as ours – that gave HP's blade servers a major power consumption edge over IBM's gear with larger configurations. It has taken IBM just a few days to fire back a retort, claiming that the HP study uses an unusual configuration and that in real world tests – you guessed it – IBM beats the snot out of HP. The vendors' aggression here underscores their two-sided domination of the blade market and intense desire to keep gaining customers in a higher-margin, largely proprietary space.
The central beef at hand comes from this paragraph in the HP report.
For a small server configuration, with non-hotplug disk drives (i.e., no hot-swappable RAID mirror) and with modest RAM requirements, there is very little difference in power consumption between the IBM BladeCenter-H and the HP BladeSystem c7000. As the server configuration increases, however, the HP blade outshines the IBM for power utilization by either 16 per cent or 27 per cent, depending on the need or lack of need for hot-swap disk drives. Clearly, IBM's decision to rely on expansion boards for system needing more memory or I/O capability hurts the BladeCenter-H's power consumption figures.
Exactly!, says IBM.
IBM reckons that the system used in the test shows "how far HP is willing to go to bend the truth."
While IBM offers Memory and I/O and Storage Expansion blades, the vendor finds that only about 10 per cent of customers pick up the expansion products. So, HP's testers have crafted "an unrealistic comparison at best," since very few customers would choose to run both of the expansion blades at the same time.
"IBM testing shows the (IBM) HS21 with the Memory and I/O Expansion Blade and Storage and Expansion Blade as using 40-50 per cent more power than a base HS21," IBM said. "If HP’s results are only showing them as having a 27 per cent advantage on this configuration, that validates IBM testing stating IBM is up to 24 per cent more energy efficient than HP."
The IBM report goes on to attack HP's air flow and cooling claims – the typical stuff at the top of customers' minds.
Stats aside, the jabs from the vendors really arise from a recent report that showed HP had grabbed the blade server revenue and shipments lead away from IBM. Last year, HP popped out an overhauled blade chassis – the c-Class system – and this product has been well received. IBM, thus far, has proved less willing to do a radical redesign of its blade chassis.
IBM's contention that you can prove just about anything by looking at isolated cases is, of course, right. The vendor's blade track record, however, doesn't exactly help it out in power consumption debates. IBM used to offer up a warning that customers not fully fill a server rack with its blades out of heat concerns and has shipped ever-larger power supplies for its blade systems on a regular basis.
The decision of whether to go with HP or IBM hinges on the quality of your data center and the software you're running, but you knew that without the help of some hired consultant. Well-cooled data centers can handle blade systems of all sizes, while less flash data centers will struggle with a rack full of systems from HP, IBM, Dell, Sun Microsystems or your name it.
The verbal sparring between HP and IBM avoids this obvious state of affairs and keeps the vendors' names in the press, which is exactly what they want. Sun and Dell seem rather hapless in the blade market, and HP and IBM would like to pick up as many sales as possible until the rivals get their acts together. (Incidentally, we hear of a Dell chassis redesign scheduled for later this year. What do you hear?)
These systems may run on so-called industry standard parts, but once you're locked into a chassis from a certain vendor, you're probably sticking with them over the long haul.