On Thu, 2003-10-16 at 14:47, Rich Smrcina wrote: > With expanded storage VM provides a multi-level paging structure. More > frequently used pages would typically live in expanded storage so that > they can be 'paged in' faster. Less frequently used pages will migrate > to disk naturally. There should be a fair amount of doc on the VM web > site that describes this in much more detail.
There's a lot on that in the recent Domino redbook SG24-7021 as well. Certainly at the amounts that were quoted this should apply. If you're going much bigger, then things get different, depending on the z/VM configuration. I hope the Linux guest with 64 MB + 128 MB swap is just pure coincidence, right? There is a myth about swap being twice as much as memory. That is completely not true. If you have performance tools that show you where you virtual disk pages are (in main, expanded or on paging dasd) then you should see that your footprint grows over time (if you do at least some swapping). This means that over time the Linux guest will be more expensive for z/VM to run. Compare that with the Linux view on swap (i.e. the amount of data on swap device as reported by vmstat - or free). You should then compare that with two virtual disks for swapping, a small one and a big one, defined with different priority (the default, but unlike what some others may think). Now small and big are vague sizing instructions, but the small one should be large enough to hold what you typically have swapped out (the number obtained from vmstat in the previous paragraph). The size of the bigger one depends on what is swapped out at peak times, and if you never swap more than the first, then you did not need the second disk at all ;-) The final version of the Domino redbook should have some more detail on this. Rob
