>>> On 11/17/2009 at 12:18 PM, Lindy Mayfield <[email protected]> >>> wrote: > I read in z/Journal that one mainframe can host 1,500 Linux servers. What > sort of mainframe can do this? How many CPU's would it take? How many CPU's > are the maximum?
As any discussion of capacity or performance requires, "it depends." You're not going to fit that much CPU-intensive workload onto any mainframe available today. You could fit that many 0-1% busy systems on a couple of IFLs. Real life workload is going to be somewhere in between (big surprise). I believe (without having found the person or persons that originated this number) that the 1,500 number came from extrapolating the "average CPU busy" of a modern Intel system to a full-blown z10 EC. As usual, marketing trumps technical details. > I also read in z/Journal that the lines between a mainframe computer (the > z10 to be specific) and a super computer are being blurred. Given that most supercomputers these days are clusters of hundreds/thousands of Intel or PPC boxes, no way. Mainframes are good, but they're not magic as we well know. > When I was at > GuideSHARE Europe two years ago (in Dresden, lovely city) they had a hardware > guy there next to a z10 with the nice green stripe down it, and he told me > that the mainframe is great for transactional processing, as always, but not > too much suited for WebSphere, Java stuff, etc. That's why they had to add > speciality engines, etc. Well, that's how I remember it. If that was indeed what he said, then he was confused. Specialty engines were introduced for sales/marketing/political reasons, not technical ones. IFLs, the original specialty engine, were created specifically so that customers could add Linux workloads to their existing mainframes, and not have the additional capacity bump up their z/OS software charges. That worked so well that zIIPs and zAAPs followed, but for z/OS only. (Linux systems don't "need" them because the pricing is different to start with.) It's not that CPs are "not well suited" for things like WAS, or DB2. It's because customers resisted buying more of them because of the increased software costs. If IBM were to replace all the zIIPs and zAAPs with standard engines, and your software costs didn't change, customers would be much better off since they would get the full use of the engines, and things would run just as well if not better. Of course, that's not going to happen, from all that we've seen o! ver the years. Mark Post ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
