We went from -3s to -9s, but not to 1/3 as many. We converted because we (a) ran out of page space, and (b) needed to fit within the limits for the CP_Owned list. We actually went to more -9s than the number of -3s that we were using. Since then, we have not had as great a load on the system as we did when we ran out of space, so comparing current performance to that max load would not be meaningful. And it will get even less meaningful when we migrate to a z10 next year.
That said, we have had no paging bottleneck and we have not filled our page space since the conversion, so the performance must be better :-) That is anecdotal rather than being based on measurements, so neither Barton nor Bit would be interested. Come to think of it, I am not very interested, either. Regards, Richard Schuh > -----Original Message----- > From: The IBM z/VM Operating System > [mailto:[email protected]] On Behalf Of O'Brien, Dennis L > Sent: Tuesday, December 01, 2009 12:32 PM > To: [email protected] > Subject: 3390-9's for paging > > Our storage architect is trying to get us to use 3390-9's > instead of 3390-3's for paging. I know that z/VM doesn't use > PAV or HyperPAV for paging, and that more devices means more > concurrent I/O's. Has anyone actually tried changing from > 3390-3's to one-third as many 3390-9's? If you have, was > there a measurable difference in performance? > > Dennis > > "I always remind people from outside our state that there's > plenty of room for all Alaska's animals -- right next to the > mashed potatoes." -- Sarah Palin >
