Tuesday, February 28, 2006

Consolidate or go for grid of cheap servers?

Most fortune-500 companies have already done some consolidation. The benefits of server consolidation have been huge - the tens or hundreds of servers, each with different patch levels and service contracts contributing to low utilizations have been replaced by a few high-end servers. Nicholas Carr, in his blog asks if the server industry will suffer as a result of higher CPU utilization (resulting either from virtualization or consolidation)?

SAP's Charles Zedlewski provides a good counterargument, why the server industry will be ok after all.

The real metrics that organizations should consider are
1. Costs per processing unit,
2. Volatility of computing demand.

In the long term, enterprises of the future will go with a limited number of highly consolidated servers running some virtualization software, yielding about 60%-70% in utilization, with additional computing power available on-demand to run complex statistical and marketing programs, provided by OEM vendors like Sun, IBM or other providers using generic hardware.

I wouldn't be too worried about any drastic reduction in computing demand, simply because firms would (hopefully) buy this additional lower cost processing capacity to make better business decisions. The only server manufacturers who would truly go away are server divisions or companies that make expensive, generic hardware (Dell, HP?).

No comments: