Chris, Setting quickdsp on for one or two primary servers is exactly what quickdsp is for. Using it indiscriminately is as Adam points out is not correcting a problem. In general, servers such as racf and tcpip that support other users should have quickdsp. A simplistic way of looking at quickdsp - If (When) performance gets really bad, does it help or hurt to dispatch this guest. If other servers have work in progress and are waiting for a database server or other server that is not dispatchable, then now you have two servers waiting to be dispatched. If TCPIP is on the eligible list, everything stops, obviously not good for performance. Evaluate shared servers in this manner.
Mike, there are many scheduler problems with linux under z/vm in that there are many reasons why linux servers do not drop from queue. At the tech conference Brian Wade pointed out that z/VM 4.4 fixed some of these, and z/VM 5.1 fixed the rest or at least the most serious of the rest. If you are on 4.4, use virtual switch instead of dedicated osa is one bypass. The problem is that if linux servers are idle but do not drop from queue, then z/VM fences their storage and will page it only as last result. So any algorithm inside CP that evaluate storage requirement is wrong if there are idle/inqueue servers. To bypass this problem in CP, SET STORBUF 300 300 300 >Date: Wed, 1 Dec 2004 12:20:21 -0600 >Reply-To: Linux on 390 Port <[EMAIL PROTECTED]> >Sender: Linux on 390 Port <[EMAIL PROTECTED]> >From: "Little, Chris" <[EMAIL PROTECTED]> > >What about a mission critical guest? We have QUICKDSP set to ON >for our primary database server. We _want_ it to work at the >expense of everything else. If things get tight, that guest >should not suffer. > >-----Original Message----- >From: Adam Thornton [mailto:[EMAIL PROTECTED] >Sent: Wednesday, December 01, 2004 10:49 AM >To: [EMAIL PROTECTED] >Subject: Re: Linux Slowdown > > >On Dec 1, 2004, at 9:59 AM, Mike Riggs wrote: > >> Should QUICKDSP be set ON for the LINUX guests under VM? >> >Not in general. > >It may help a particular guest, but it does so at the expense of >everyone else, and basically you're patching a symptom rather >than fixing the underlying problem. > >Adam > "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
