> IBM states (AFAIRC) somewhere that the CPW is calculated on the "maxed" out > configuration. This is hard to believe: do they really mean (for their biggest box) > that CPW is determined for a box with 128GB RAM and 19TB DASD? Believe it. Not every configuration is measured, of course - too expensive. Some are extrapolated. If you will look at official CPW tables, those which were calculated (not measured) are explicitly marked so. The largest configuration is, however, measured. > With most other platforms the various > benchmarks (as you point out) are more or less public, so some validation > of the numbers is possible. The CPW is much harder to get a handle on. I am curious how can you validate, say TPC-C benchmark for maxed out Sun Enterprise configuration - cost of such benchmarks is easily in millions. I also want to say, that normally vendors are not interested in benchmarking small comfigurations. For example, you can find TPC-C figures for large Compaq servers, but do you know what is the TPC-C rating of a Compaq box you buy in a shop on a corner ? Besides, CPW was never intended for comparison with oter platforms. For this, IBM publishes lots of public benchmarks (TPC family, different kinds of ERP, Java, Lotus etc. etc). CPW was meant only to differentiate boxes in the same family, so I guess in a *legal* sense, CPW numbers are a kind of arbitrary. Alexei Pytel - speaking only for myself "Leif Svalgaard" <email@example.com> To: <firstname.lastname@example.org> Sent by: cc: midrange-l-admin@mi Subject: Re: Understanding iSeries Performance drange.com Positioning 09/07/2001 12:10 PM Please respond to midrange-l From: Nathan M. Andelin <email@example.com> > I'd like to gain a better understanding of the benchmarks that IBM uses to > calculate the relative CPW ratings of their boxes. I think you'll have an uphill battle here. IBM states (AFAIRC) somewhere that the CPW is calculated on the "maxed" out configuration. This is hard to believe: do they really mean (for their biggest box) that CPW is determined for a box with 128GB RAM and 19TB DASD? Since boxes may differ wildly in terms of (say) number of disk arms, it seems hard to compare just based on processor/feature codes. Our f(r)iend CFINT never seems to soak up I/O, so how can "the knee of the curve" be meaningfully defined? Remember that CFINT will begin to kick in after the knee, but I never see it soak up anything but CPU cycles. Seems to me that there is some voodoo going on. Some rather arbitrary numbers being floated around. With most other platforms the various benchmarks (as you point out) are more or less public, so some validation of the numbers is possible. The CPW is much harder to get a handle on. _______________________________________________ This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list To post a message email: MIDRANGE-L@midrange.com To subscribe, unsubscribe, or change list options, visit: http://lists.midrange.com/cgi-bin/listinfo/midrange-l or email: MIDRANGE-Lfirstname.lastname@example.org Before posting, please take a moment to review the archives at http://archive.midrange.com/midrange-l.
This mailing list archive is Copyright 1997-2014 by MIDRANGE dot COM and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available here. If you have questions about this, please contact