One of the problems when competing with mainframes (yes, there are still customers thinking about such things like Linux on a Mainframe ... i have my opinion ... won't repeat it here.) is the point that there are only a few public benchmark results.

So the recent Peoplesoft Payroll benchmark result is really interesting. It's for the extra-large dataset so the numbers aren't comparable to the ones i wrote about some time ago.

A M5000 was able to to process the model in 50.11 minutes. The system ran at 50% utilisation. A HP 8-way Itanium 2 rx7640 had to work for 96.17 minutes to process the same model.Both worked at 32 streams.

But the comparison isn't the interesting point about. There is another interesting benchmark: There is an benchmark for the same model size with the IBM z10 EC 2097-709 needed 58.96 minutes for the same task and ran at 70% to 80% utilisation. Doesn't sound that much of a difference but the used IBM mainframe has a list price of $6,141,671 (according to this list). I have to add that the mainframe ran at 8 streams instead of 32, but that's sound to be owed to the fact that the application ran on a 9-way system (8 ways for Peoplesoft, 1 way for housekeeping) in the system.

For $6m you could buy a lot of M5000, you could buy several M8000 or at least a few M9000. No matter how you turn it, the performance of a mainframe is really expensive. However, in my personal opinion mainframes aren't purchased for performance or features .... it's all about risk mitigation and the problem that you have to search for a shovel to ask someone who is still able to support the code for certain business processes. By the way ... when i'm old i think i will make a living out of the fact i'm relatively fluent in COBOL

Disclosure Statement: Oracle PeopleSoft Payroll 9.0 benchmark, Sun SPARC Enterprise M5000 (8 2.53GHz SPARC64 VII) 50.11 min, IBM z10 (9 gen1) 58.96 min, HP rx7640 (8 1.6GHz Itanium2) 96.17 min,, results 6/3/2010.

Read More about [Expensive non-performance...