>>> On Wed, Feb 13, 2008 at 7:58 PM, in message <[EMAIL PROTECTED]>, Adam Thornton <[EMAIL PROTECTED]> wrote: -snip- > I like defining two (assuming that my real load is less than two full > engines' worth) or as-many-as-I-expect-to-be-able-to-use-if-more-than- > two. > > Basically, this is to give my workload the best chance of parallel > dispatch. If I only have one virtual processor, I only have one thing > *really* going on at a time in my Linux guest, even if the actual work > is hopping around real processors. I don't have any hard data that > tells me this is really working, though. Does anyone?
According to Rob, this mainly just increases the likelihood that your guest won't drop out of queue, and the z/VM dispatcher will decide it's batch workload, not interactive. I removed any extraneous processors from my guests as a result of that conversation. Mark Post