Server Benchmark Gamesmanship Is Perilous For The Players, But Good For The Buyers

Last month, Fujitsu Computer Systems was able to boast that its server had achieved the best price-performance ratio on a <a href="http://www.tpc.org/">Transaction Processing Performance Council (TCP)</a> benchmark. Three weeks later, a Dell unit took the top spot. Watching the show may be dizzying, but it offers non-obvious advantages for server buyers.

Lamont Wood, Contributor

August 28, 2008

2 Min Read
InformationWeek logo in a gray background | InformationWeek

Last month, Fujitsu Computer Systems was able to boast that its server had achieved the best price-performance ratio on a Transaction Processing Performance Council (TCP) benchmark. Three weeks later, a Dell unit took the top spot. Watching the show may be dizzying, but it offers non-obvious advantages for server buyers.The latest back-and-forth action concerned benchmark E of the TCP (TCP-E) which simulates the online transaction processing (OLTP) workload of a brokerage house. Fujitsu sent out a press release in late July boasting of achieving a price-performance ratio of $523.49 per tpse (transactions-per-second on the E benchmark), which was the best so far. The previous titlist had been the Dell PowerEdge 2900 III with a price-performance ratio of 694.08, set in April.

Then, less than three weeks after the press release went out, the title returned to Dell, when the Dell PowerEdge R900 achieved a price-performance ratio of $500.55.

Of the last ten leaders in this price-performance war, six have been Dell systems, three have been IBM system, and one was a Fujitsu system.

But while all this one-up-manship is doubtless great sport for the participants, the advantages for a buyer are less obviousbut are there if you look. Basically, the TCP-E benchmark price (used to figure the price-performance ratio) includes everything needed to make the simulated brokerage application work, including all the storage and the necessary software, all with a three-year maintenance contract. In the Fujitsu simulation, the storage costs three times more than the database server, the software costs as much as the storageand let's not forget $21.25 for cabling and $15.40 each for optical mice. The total was $161,181.

This use of such a large niche application probably makes the results uninteresting to a small office manager trying to make a purchase decisionexcept that the component prices used in the calculation are realistic, with expected discounts. Potential buyers could use these as examples of what would be needed, and what the components are likely to cost. When looking at the results, simply click the name of the system, and in the resulting drill-down screen click the link to the supporting documentation.

Be prepared to be educated.

Go to the bMighty Server How-To Center

About the Author

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like


More Insights