Blades Of Glory

Methodology
We asked each vendor to provide us with a blade system based on industry-standard CPUs and components, with no more than four servers each. We ran Microsoft Windows Server 2008 on each server. Other than meeting our routine electrical requirements, we then left it up to each vendor to configure a system that would take each metric in our methodology into consideration. A vendor might choose to focus all efforts on raw throughput performance, but the configured price could put it at a disadvantage over other systems. Or, a vendor could configure a system that was extremely easy to deploy and manage but take a hit from reviewers when it came time to evaluate performance.

Reviewers first used Geekbench 2.0 to test performance on each server in each blade system. We also included PassMark performance benchmarking, which we, in the end, decided might be more appropriate for 64-bit testing. Noise measurements were taken at startup and idling with a decibel meter. We took heat measurements after one hour of runtime and wattage measurements at startup and idling. Reviewers also judged failover capability, ease of deployment, management features and considered configured price, channel programs and VAR profit potential.

In the end, our group of reviewers wound up agreeing that Super Micro provided the best value based on a combination of all the factors we deemed relevant. IBM, we agreed, provided a stellar and valuable solution and a chassis design that appears second to none. Dell is very new to the blade server space, and the electrical limitations of our lab provided a minor challenge. But it was terrific from a performance standpoint and, if Dell is serious about working with the channel, could easily take a little share away from HP.

Next: Super Micro OfficeBlade