Barton might have some hard numbers, since Velocity has been pushing XIP for quite a while now.
The benefit is academically obvious. To get your own measurement, you'll need a multiplicity of guests and you'll want to measure VM host load (esp paging) and Linux guest throughput (esp starting execution of programs from XIP space). From the Linux side, execution should be instantaneous. It may help to squeeze memory on both VM and Linux (for the sake of your test), so you may want to run it in a memory constrained test LPAR. -- R; <>< On Tue, Aug 10, 2010 at 17:13, Leland Lucius <[email protected]> wrote: > So, I've got an itch to fiddle around with XIP, but I can't figure out how > to measure the benefit I might be able to expect or, if I just do it, how to > figure out if I gained anything other than hassles. > > I've played around with it a little using a 512MB DCSS populated with /lib, > /lib64, /usr/lib, and /usr/lib64, but I didn't really "see" any changes to > working set sizes in Perf Toolkit. Of course, that's not really saying much > since most of the numbers in Perf Toolkit are worse than Greek to me...more > like Goa'uld. > > I know there's probably no easy answer. I also know that just arbitrarily > throwing those directories in there wasn't the best use of DCSS space and, > if I continue to fiddle, will put more effort into identifying appropriate > candidates. (I'm thinking that /proc/*/maps would be a good source for > finding the candidates.) > > Thanks, > > Leland > >
