On 3/23/07, Stracka, James (GTI) <[EMAIL PROTECTED]> wrote:
And it is considered bad practice to give a guest more virtual CPUs than real CPUs.
I would rather say "more virtual CPUs than the number of real CPUs you expect to work for that guest at some moment" The number of real CPUs is a maximum indeed but that rule was to say you should not define virtual 2-way guests if you have just one real CPU. What you want to avoid is that the guest OS dispatches work on a virtual CPU and then find that that virtual CPU does not get dispatched real soon. With 100 Linux virtual machines on a real 10-way, I think it would also be unwise to run virtual 10-way servers. Your performance monitor can show you how many virtual CPUs compete for real CPU cycles and how likely your guest would be able to run on N at one time. Much easier is not to define more than the guest can eat (or than you want to let it eat). My understanding of VSE is that a single partition will still just use a single CPU, but multiple active partitions at the same time can use different virtual CPUs. So running primarily one partition, you do not speed that up with more virtual CPUs, even when they get dispatched on z/VM. Rob -- Rob van der Heij Velocity Software, Inc http://velocitysoftware.com/
