-----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Staller, Allan Sent: Tuesday, November 17, 2009 1:05 PM To: [email protected] Subject: Re: Mainframe running 1,500 Linux servers?
IIRC, an experiment was done a few years back to see how many Linux images could be started in an IFL using z/VM. IIRC, HW resources ran out before the software architecture limits were reached. ISTR > 32k images running concurrently (no useful work, just how many can be started). <SNIPPAGE> Some months ago I was reading some tuning information. One of the things that was pointed out, for z Linux under VM: LIMIT THE MEMORY. If you have n Linux guests, and all have 6GB available to them, they will all go for the 6GB. This could cause your z/VM system to have a serious AUX shortage. The reason given for this, and I see it happening on my own personal machines, is that Linux will load all the stuff that it needs and then will use as much memory as it can for caching and then it will start using its swap (paging) files. So as I load applications, I notice the cache numbers starting to drop. Once I get past a certain number, SWAP starts being used. Now, when I took one of those machines and added 2GB of memory, Linux immediately gobbled it with a large area used for cache. Given this, if your z Linux only requires 1GB, only give it 1GB and let it use its swap files. The one article I read on this said that when they did this, they gained back both available CPU and AUX space. Regards, Steve Thompson -- Opinions expressed by this poster do not necessarily reflect those of poster's employer -- ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
