Tim said > Re: Bruno's point about VMware, <<>> you can't overpower X86's > limitations with processor horsepower, and modern IFLs are pretty decent > in sheer number crunching ability anyway.
I am not sure what you're saying here Tim. There's a wide spectrum of processor "horsepower" available within the X86 line, just as there is on z. However, if you want raw horsepower, you can get a lot more of that from mid to high end X86 architecture - at any price point - than you can from the high end of z. High end Intel (and equivalent) chips blow the doors off z for compute power. So do most of the RISC chips like the IBM Power architecture. If you're simply saying that IFLs are "powerful enough" for most users then that is certainly true for some. It may or may be true in general, depending on the user and the workload. I would not make a presumption either way. It would be fair to say we have had some (ahem) "customer resistance" to the idea of offloading management stuff to zLinux. LinTEL is seen as more attractive because of the price/performance argument. And note that I am not taking a position on whether those views have merit - just reporting they are fairly prevalent out in customer land. > z/VM and z/Architecture are very > good at what they do and, while VMware is quite a step forward, it still > has very real limitations. Even with VMware we're just not seeing > processor utilizations going up all that much in real world experience (as > evidenced perhaps by Bruno's 240 servers :-)). This is an area where there are implicit assumptions of goodness with respect to cpu utilization that may not have any validity in the real world. If your cpu is really expensive (like a hundred grand expensive) then you're going to want to wring every last bit of productivity from it that you can. That's how we got to where we have with the 360 line and its successors. On the other hand, if the cycles are absurdly cheap, you -might- still care about their utilization but only if there were other costs that were geared to the number of instances in use. That is the driving factor in VMWARE (or XEN) virtualization efforts in the open systems market place. It isn't that anyone cares whether the processor is 1% or 10% or 50% busy. It is that it costs more in people, power, heat or floor space to have those things ganged up on the floor. The more ticklish question is whether or not they're still more cost effective, even with the sprawl issue. And there again, I am somewhat agnostic. The case can be made on either side. <<>> > No question, though, that in many enterprises both technologies should > play a role. It depends on the workload of course: blade servers, AIX LPARs, > and z/VM all exist because they have different characteristics That's really the whole issue in a nutshell. CC ---------------------------------------------------------------------- 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
