donb...@gmail.com (Don Williams) writes: > It does not actually matter whether their systems are "open" or not; just > different from IBM. Yes, the total cost is less expensive, because they are > smaller. Yes, the cost per transaction is higher. However, the cost per > transaction may be out weighted by other factors. Regardless, they'll work > hard on making it cheaper per transaction. Will they succeed? May be, maybe > not; but I think they have a good shot it. What I'm really saying is that > there is more than one way to get great results. IBM is not guaranteed to be > the market leader. IBM will have to fight to keep the lead and it will not > be an easy battle. Of course, IBM will fight for it.
re: http://www.garlic.com/~lynn/2013h.html#39 Why does IBM keep saying things like this: http://www.garlic.com/~lynn/2013h.html#44 Why does IBM keep saying things like this: there are lots of industry standard benchmarks ... which require published results of transactions per time as well as total cost per transaction (and for DBMS, they have to demonstrate conformance with ACID properties). long ago and far away, there were mainframe results for these benchmarks ... but not for a long time. it isn't a vendor issue since ibm does do these benchmarks for the other platforms it sells. example http://www.tpc.org/ tpcc (both transaction performance and price/transaction) http://www.tpc.org/tpcc/results/tpcc_perf_results.asp a handle on part of mainframe total cost of ownership is IBM's mainframe revenue ... on the average. IBM's mainframe division earns total of $6.25 for every dollar revenue from processor ... aka a customer paying $28M for a 80-processor max configured z196 ... would avg. total payments to IBM of $175M (which is just the start for running mainframe datacenter). 80-processor max configured z196 is rated at 50BIPS ($175M/50 - $3.5M/BIPS). IBM's peak I/O benchmark for z196 is 2M IOPS using 104 FICON channels ... with 14 system assist processors. Peak SAPs is 2.2M SSCH/sec running all 14 at 100% busy, but recommendation is keep SAPs to 70% or less (1.5M SSCH/sec). IBM base list price for e5-2600 blade is $1815 and e5-2600 have ratings of 527BIPS or $3.44/BIPS (factor of million times less than z196). Recent FCS (FICON is heavyweight protocol layer on top of FCS that drastically cuts the native throughput) for e5-2600 claims over million IOPS (i.e. two such FCS would have higher throughput than 104 FICON) disclaimer: I worked with Jim Gray at IBM before he left for Tandem ... and his work to create tpc: http://www.tpc.org/information/who/gray.asp -- 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