Please note that I don't have enough mainframe experience to comment on Linux running on that platform, but I'm going to shoot my mouth off anyway. That said, and assuming you don't consider me an "authority" (I'm only a "stand up philosopher", after all):
I once, a few years ago, wrote a note to a fellow about what a mainframe *is* that differentiates it from other platforms. I was surprised some time ago to discover that my "naive" description is still quoted. So... to recap http://mainframes.com/whatis.htm : 1) Maximum single-thread performance 2) Maximum I/O Connectivty. 3) Maximum I/O Bandwidth. I would now revise (in light of Appendix "A" of "Linux for the S/390", SG24-4987) 1) Maximum [RELIABLE] single-thread performance So an Intel (or a PPC) _may_ smoke a zSeries doing number crunching *but* loses that advantage should a single processor develop a fault. A zSeries, AFAIK (and I'm no expert there, being more of a pSeries guy) is *made* to be serviced whule still runningi, dealing w/ CPUs, and memory directly. (Sometime I wonder but it _is_ amazing to me.) I don't have any access to top-end pSeries boxes but some of the boxes I've seen can have power supplies and the like swapped while it's running. So mainframes are about fault detection and recovery to ensure ACCURATE results. Especially when doing Division... :-) And "Appendix A" is important reading for those wondering why you'd *choose* to use a mainframe. -------------------- John R. Campbell, Speaker to Machines (GNUrd) {813-356|697}-5322 Adsumo ergo raptus sum IBM Certified: IBM AIX 4.3 System Administration, System Support http://packrat.tampa.ibmus2.ibm.com/~soupjrc/ Backup: Toby Schmeling {813-356|697}-5233 David Boyes <[EMAIL PROTECTED] To: [EMAIL PROTECTED] e.net> cc: Sent by: Linux on Subject: Re: [LINUX-390] InfoWorld Article - Microsoft Benchmarks Step Up Linux 390 Port Assault <[EMAIL PROTECTED] IST.EDU> 09/05/2003 02:11 PM Please respond to Linux on 390 Port > He obviuosly does not understand the power of the > mainframe. IBM has always talked about throughput > rather than "performance" - elapsed time of any given > transaction and throughput. Oh, I'd think he understands all too well. Most of the industry "performance" benchmarks measure things that mainframes aren't good at. It's like measuring the duration of your next trip to the grocery store in angstroms per cubic kilometer. The measurements are meaningless, or are misleading in ways that don't benefit IBM, so there's no point in expending lots of resources to compare something that won't tell you anything positive. This is long-standing IBM policy -- nothing new to see here.
