david.dev...@sse.com (David Devine) writes: > Looking at processor and software costs in isolation doesnt tell the whole > story. > > Yes, software cost are a big chunk, but doesnt Microsoft charge like a > Rhino for each Windows licence? > > What would you attach your E5-2600 blade to and using what? fibre or > ethernet? whose disk systems? tape for backup? > how resilient is it? how many staff would it take to manage? > > The elephant in the room is reliability. > > Z/series and associated kit is solid and dependable (baring a few > exceptions) having grown ergonomically over 50 years. > > How much down time do you get from windows or Unix farms? > Would you risk running your key billing platform's on flaky kit? you > can't send your bills out you can't get your money in.
re: http://www.garlic.com/~lynn/2013b.html#5 mainframe "selling" points http://www.garlic.com/~lynn/2013b.html#6 mainframe "selling" points one of the things recently mentioned is that the large brand vendors (HP, DELL, IBM, etc) are no longer the major consumer of i86 server chips ... that the large cloud operators (both public and private) are ordering chips directly and building out mega-datacenters with several hundred thousand blade configurations (and millions of cores). There is lot of similarity between the millions of core supercomputers and the millions of core cloud operators. Periodic press is also that the cloud operators are building their own blades at 1/3rd the price of brand name blades (bring cost/BIP into the $1 range). The big cloud operators have also done extensive work on choice of system components to optimize the price/reliability, price/maintenance, price/adminstration, etc issues. With the enormous drop in processing costs ... the large cloud operators have also turned their attention to all the other operating costs that are now starting to dominate ... maintenance, power, cooling, administration, etc. With hundreds of thousands of blades and millions of cores ... they have done an enormous amount of work optimizing all these other costs. For the majority of the e5-2600 blades out there in the large cloud operations (public & private), there are running various flavors of Linux (significanlty reducing those costs) and have processes where a relatively few people are able to operate a mega-datacenter with millions of cores. With the lots of on-demand characteristic in many cloud operations ... they are behind pushing for almost zero power&cooling while idle and able to be brought up to full operation nearly instantaneously. Guess is that any one of the numerous mega-datacenters around the world has more processing power than the aggregate of all mainframes in the world today. For other drift ... real CKD disks haven't been manufactored for decades ... they are all done with simulation using the same disks that are used by e5-2600 systems. Also as mentioned upthread ... native FCS has enormously better throughput than when FICON is layered on top of FCS ... aka peak z196 I/O benchmark with 104 FICON channels at 2.2M IOPS compared to announcement of a *single* native FCS for e5-2600 capable of over 1M IOPS. -- 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
