What David Rock said. And I don't think VMware is at all to blame. I *suspect* that virtualization of Pentium is hard to do. zSeries has more than three decades of hardware and software adjusting for each other to do virtualization very efficiently.
I use VMware all the time. Not "continuously" anymore because of the performance and stability hits, but still several times a week and often daily. I love it! But compared to z/VM: there is little to discuss. z/VM simply blows VMware away. "Insertion loss" is a good metric for comparing hypervisors. I don't have hard numbers to give you, but a ball-park estimate would be something like 10-30% for VMware, but like 0.5 to 5% for z/VM. And while I don't know if it is true today, there have been times in the history of IBM's VM where a guest would see "greater than unity" throughput, that is, the guest ran *faster* on VM than native. (That was the exception even then, but it was and is laudable.) So VMware is great, but z/VM is better, and a lot of the difference is because the zSeries hardware and z/VM are both more grown up and have grown up TOGETHER. VMware is new, and INTeL has never addressed virtualization to the extreme of virtual machines. (I do not count the "virtual 8086" that they've done for a decade. For "VM" of any flavor, you need a CPU that can virtualize itself, not just virtualize a predecessor.) This is not meant to slam INTeL. It is what it is and it does well for certain jobs. You asked: > Has anyone done a comparison of the virtualization capabilities and cost > effectivness of z/Linux w/z/VM to using VMware on Intel > and if so, what conclusions did you draw ? It depends entirely on what kind of job you need to do. What's great about zSeries and INTeL is that you can run Linux on either and you can have virtual machines on either. Would that that were true of PA-RISC, SPARC, PPC, ... etc, etc. :-(
