On 8/31/07, Rich Greenberg <[EMAIL PROTECTED]> wrote: > } pitfall: there is (currently) no performance data. > } > Well, then Rich will have the oppurtunity to provide some. :-)
But the reason we do not have performance data is that there is really no instrumentation... For z/VM disks on FCP (aka edevice) there is some made-up numbers from CP, so you would be able to tell that you have bad disk I/O response, but little help to determine why that is (other than blame those folks of the SAN). For the Linux disk I/O on FCP, you have no useful instrumentation from VM or channel. There is some experimental code to measure things in Linux, but it would be even harder to make sense of such numbers with Linux on z/VM. Obviously we did review performance data from z/VM systems using FCP for their Linux DASD only, as well as for the entire z/VM system. I don't recall a situation where the FCP disk I/O was determined to be the bottleneck. That does not mean FCP never is a problem. The question is whether you have enough ammunition to point a finger when you do have disk I/O problems. Rob -- Rob van der Heij Velocity Software, Inc http://velocitysoftware.com/