Look at the user seek report - it shows you which devices have I/O by minidisk, by user. Find out where the I/O is going. Then look at the dasd report (ESADSD2) at those device(s), is it connect time, pend time, discconnect time or queueing that shows up on the high side? If you mix escon and ficon channels, there are cases where the I/O processor runs at 100% - and pend time increases. Or if on escon channels, i/o will be slow anyway. If high queueing, then maybe PAV could help or other buffering. Or as it is database, one installation took a database that one process was taking days, added an index and process took minutes. if high connect time, shrinking the machine would have a positive impact. or if you swap to disk, you would have i/o wait, in which case increase machine size.... Is it oracle? does the sga fit in page cache? what is sga size? so many choices, so little time. ship your performance report to velocity software if you like..
>Date: Thu, 17 Aug 2006 13:02:55 -0400 >Sender: Linux on 390 Port <[email protected]> >From: Richard Troth <[EMAIL PROTECTED]> > > >----- Original Message ----- >From: Dave Jones [EMAIL PROTECTED] >Sent: 08/17/2006 12:24 PM >To: [email protected] >Subject: Re: I/O wait times -vs- Linux memory > >Rick, what kind of Linux i/o are you doing? To real 3390s? To SHARKs >looking like a 3390? To some sort of SCSI attached disk? > >DJ > >Richard Troth wrote: >> We're looking at some high I/O wait times for a certain database. One >> engineer has suggested that we add memory to Linux to speed things up. I >> don't see it. Does adding RAM to Linux help its I/O throughput? >> >> We've put a lot of effort into sizing our Linux guests (all hosted by >> z/VM 5.2) to that magical point between guest-level paging and excess >> guest memory use, so some of us resist the thought of adding >> real-to-Linux memory. I seems obvious that adding RAM would let any >> access method do more buffering, but we want to avoid some of that >> buffering because Linux gets carried away with it, correct? >> >> Thoughts? Thanks! >> >> -- R; "If you can't measure it, I'm Just NOT interested!"(tm) /************************************************************/ Barton Robinson - CBW Internet: [EMAIL PROTECTED] Velocity Software, Inc Mailing Address: 196-D Castro Street P.O. Box 390640 Mountain View, CA 94041 Mountain View, CA 94039-0640 VM Performance Hotline: 650-964-8867 Fax: 650-964-9012 Web Page: WWW.VELOCITY-SOFTWARE.COM /************************************************************/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
