On Thu, 10 Dec 2015 14:05:26 -0800, Barton Robinson wrote: > If anybody in endicott is listening, ONE REQUEST. The current diagnose > used by the CMM function to release a page actually causes z/VM to read > the page in from the paging subsystem before releasing it. So release > 100,000 pages that are on the paging subsystem means 100,000 page reads > from disk. VERY BAD. If you discard a page that is on paging disk, > discard it, don't read it and then discard it.
Sorry, Barton, but this is simply factually incorrect, and would certainly be APARable if it were so. We do not read data pages from DASD during any virtual storage release operation. However, it is true that releasing them may require paging I/O reads to read the PGMBKs which represent the megabytes containing the released pages. The ratio of PGMBKs to pages represented is 8K to 1M (256 pages), but the actual ratio we need to read can be substantially higher depending on the distribution - if only one page per megabyte is being released, we may have to read one 8K PGMBK for each 4K page being released. If you have evidence that actual data pages are being read during any release operation (diag x'10' or otherwise), please open a PMR and send in your evidence. (re-sending with corrected subject, sorry) - Bill Holder, z/VM Development, IBM "Much that once was is lost, for none now live who remember it." - Galadriel "He that breaks a thing to find out what it is has left the path of wisdom" - Gandalf (JRR Tolkein) ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/