All, Another element of paging that has not been referenced is the ability to handle all of the swap set size in parallel. If the swap set size is 120 pages then the old practice was to have at least four LOCALS so each thirty page block of pages could be swapped-in in parallel. While swapping, like paging, is not as prevalent as it once was I'm wondering if the swap set size is still one of the principal guidelines for the number of locals that should be defined.
For HDS VSP customers there is a new facility called MF-HDP that allows for some very wide striping of volumes across the available spindles. If you are using or plan to use MF-HDP then LOCALS would be very good candidates for HDP pool volumes. You can allocate a 3390-3, 9, 27, 54 or A as a virtual volume within the pool, but initially the space you use will be a 672 track block that contains the volume label, VTOC, Index and VVDS. Then when you define and format you LOCAL you will only use space equal to the size of the page dataset rounded up to 672 tracks. So if you want to allocate a 3390-54 for your locals, but only make them 5000 CYLS in size you should go for it, because the 60020 CYLS of empty space won't actually use any space in the HDP pool. If you handled this concept on Iceberg and the RVA then you're well on your way to wrapping your head around this with MF-HDP? The other advantage of this is the wide striping I mentioned. Each 30 page set of contiguous slots will be within the same page, but the page is striped across the underlying parity group disks. There won't be much advantage for block page-in for each set of thirty pages, but you don't have to worry about hand placing all your locals across the parity groups. The wide striping will uniformly distribute all the page datasets across all the underlying parity groups and disks. If you have 128 parity groups of R6 6D+2P then the read miss and destage activity of your locals, no matter how many, will be uniformly spread across 1024 disk drives. Ignoring UCB constraints, it kind of makes minimizing the number of locals an academic exercise. If you think you need eight locals then allocate sixteen that are half the size on 3390-A. You will still only use the same amount of disk space. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@bama.ua.edu] On Behalf Of > Barbara Nitz > Sent: Wednesday, February 01, 2012 9:13 PM > To: IBM-MAIN@bama.ua.edu > Subject: Re: [IBM-MAIN] Very Lage Page Datasets (was ASM and HiperPAV) > > >So if I have 5 3390-27 locals and they are all equally used at 50%, the > >algorithms (CPU usage, not I/O) are going to pick one of them, then do > >the page outs. That paging will find contiguous slots and should be > efficient. > > > >BTW, this is just an example, we still try to keep our 3390-27 local > >usage at 30% just like we always did with smaller local page datasets > >in the past. > > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: INFO IBM-MAIN
