IBM Challenges Competitors in New Benchmark

[MCWebsite.Staff]

** This thread discusses the article: IBM Challenges Competitors in New Benchmark **
** This thread discusses the Content article: IBM Challenges Competitors in New Benchmark **
0

[J.Pluta]

IBM Challenges Competitors in New Benchmark

** This thread discusses the article: IBM Challenges Competitors in New Benchmark **
Lee, this is perhaps the most cogent discussion I have seen to date of the great quagmire that is benchmarking. We all know that you can tune your machine to run a given benchmark as fast as possible; the question is how that relates to real world transaction processing. Personally, I'm glad that there is now a benchmark using a real enterprise application like SAP. I've always known that once you got out of the theoretical world and into an application that actually did real work that the i5 architecure would prove itself. It's nice to see the empirical knowledge that is so obvious to us in the midrange world is finally finding an expression in definable, repeatable test models. But your points about utilization and price per user are right on the mark; until we properly take those into account, decision makers will still be getting incomplete data. I don't know what the answer is, but I agree that getting some more real world tests into the public domain is the first step. Anyway, thanks for taking the time to extract some meaningful information from the sea of statistical data. Great job! Joe

[lkroon]

IBM Challenges Competitors in New Benchmark

** This thread discusses the article: IBM Challenges Competitors in New Benchmark **
Thank you for your kind words, Joe. As I implied in the article, I'm frustrated by the inability of industry benchmarks to reflect real-world price/performance. I'm going to talk with Al Zollar about this the next time that I see him; perhaps we can start a movement within IBM to create an open, industry-standard benchmark that tests the ability of different platforms to manage multiple workloads in a partitioned environment. I am convinced that such a benchmark would demonstrate that the effective price/performance of the iSeries is superior. Of course, it is doubtful that Intel server vendors would agree to publish how their servers perform on such a benchmark. That is why most benchmarking will probably continue to be more akin to benchmarketing (sigh). The truth is hard to come by, ain't it?