sme...@gmu.edu (Seymour J Metz) writes: > Of course, IBM does not claim that those numbers reflect a Meaningless > Indicator of Processor Speed (MIPS), but rather a well defined (LSPR > ITR) benchmark.
Jim Gray was one of the primary behind people original SQL/RDBMS, System/R ... and then left IBM Research (trying to palm off bunch of stuff on me) for Tandem, then DEC, then microsoft. One of the things he did starting while at Tandem was standardized DBMS trasnaction benchmarks http://www.tpc.org/information/who/gray.asp https://jimgray.azurewebsites.net/ numbers per system, numbers per total system cost $$$, and more recently numbers per power. both cluster supercomputer(grid) and cloud megadatacenter so significantly drop system costs (claim for over a decade that they assemble blade systems for 1/3rd cost of brand name systems) ... that power/cooling was becoming increasingly major part of total-cost-of-ownership. I continued to see some mainframe industry standard TPC numbers up to some time last decade ... but haven't found anything since then. past published numbers of mainframe "MIPS" (now BIPS) z900, 16 processors, 2.5BIPS (156MIPS/proc), Dec2000 z990, 32 processors, 9BIPS, (281MIPS/proc), 2003 z9, 54 processors, 18BIPS (333MIPS/proc), July2005 z10, 64 processors, 30BIPS (469MIPS/proc), Feb2008 z196, 80 processors, 50BIPS (625MIPS/proc), Jul2010 EC12, 101 processors, 75BIPS (743MIPS/proc), Aug2012 z13, 140 processors, 100BIPS (710MIPS/proc), Jan2015 z14, 170 processors, 150BIPS (862MIPS/proc), Aug2017 2010 was when they published z196 "peak I/O" benchmark getting 2M IOPS using 104 FICON (protocol running over industry fibre channel standard). This was also about the same time that a fibre channel was announced for E5-2600 blade claiming over million IOPS (for single fibre channel, two such having higher throughput than 104 FICON running over 104 fibre channel). At the time, E5-2600 blades had benchmarks of 400BIPS-530BIPS (depending on model, industry standard is number of iterations compared to 370/158-3 assumed to be 1MIPS) ... and at the time IBM base list price for E5-2600 blade was $1815 (about $3/BIPS) compared to $30M for max configured z196 (or about $600,000/BIPS, not including devices, software and services). It was not long later that server chip makers announce they were shipping over half their product directly to cloud megadatacenters (where they assemble for 1/3rd the cost of brand named servers, 1/3rd IBM's $3/BIPS is $1/BIPS), and IBM sells off its server business. trivia: cache miss, memory latency ... when measured in number of processor cycles ... is compareable to 60s disk I/O latency when measured in count of 60s processor cycles. z196 claim is over half the z10->196 per processor improvement (469MIPS to 625MIPS) is the introduction of memory latency compensating technology (that have been in other platforms for decades), out-of-order execution, hyperthreading, branch prediction, etc. ... sort of hardware equivalent to 60s multitasking. FICON trivia: 1980s, STL was bursting at the seams and they were moving 300 people from the IMS group to offsite bldg (with dataprocessing back to STL datacenter). They tried "remote" 3270s ... but found human factors horrible compared to local channel-attached controllers in the bldg. I get con'ed into doing channel-extender support allowing local channel-attached controllers to be placed at the offsite bldg (and don't see any difference in human factors between local and offiste). The hardware vendor tries to get IBM approval to ship my support, but a group in POK playing with some serial stuff, get it vetoed because they were afraid that it would make it harder to ship their stuff. In 1988, I'm asked to help LLNL (national lab) standardize some serial stuff they are playing with, which quickly becomes fibre channel standard (including some stuff I did in 1980). The POK people finally get their stuff released in 1990 with ES/9000 as ESCON when it is already obsolete. Later some POK people become involved in fibre channel standard and define a heavy-weight protocol that drastically reduces the native I/O throughput that eventually ships as FICON. -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN