Terry wrote, "it is imperative that you have a robust paging subsystem"
I do not claim great knowledge in z/VM performance matters. "Where ignorance is bliss, tis folly to be wise" (Thomas Gray) Does z/VM, unlike z/OS, still need to do a lot of paging? In the z/OS world paging has become virtually(no pun intended :-)) non-existent. So what's all this fuss about paging in z/VM. True, you still need enough page space defined on disk to back storage. But after that there should be little to nil paging going on at least in the z/OS, MVS world. I know years ago there were problems with the handshaking between MVS and VM. The problem being it was virtually :-) non-existent, unlike VSE which had the VM/VSE Feature. So that whenever an MVS address space got a page fault, the entire MVS virtual machine was swapped out. But, I thought, that was addressed :-) years ago and today z/VM is aware of such things. So, then why the need for a robust paging subsystem. It all kinda leaves you "starry-eyed and vaguely discontented". "Martin, Terry R. (CMS/CTR) (CTR)" <[email protected]> Sent by: The IBM z/VM Operating System <[email protected]> 09/16/2010 09:50 PM Please respond to The IBM z/VM Operating System <[email protected]> To [email protected] cc Subject Re: CP unresponsive on certain guests I would just add that for any significant number of z/Linux guests it is imperative that you have a robust paging subsystem and enough spool space to hold CP dumps as well as other spool files. Having a well configured paging subsystem will greatly reduce the possibility of issues when your system has a high paging rate due to the over commitment of Available Real Memory versus the total memory of all users running in the LPAR combined. The higher this over commitment ratio the more paging is introduced. The good news is as long as you have a sufficient paging subsystem z/VM handles paging nicely even at relatively high rates. So over commitment of real memory in z/VM is good just keep an eye out so that the ratio stays within reason. Thank You, Terry Martin Lockheed Martin - Citic z/OS and z/VM Performance Tuning and Operating Systems Support Office - 443 348-2102 Cell - 443 632-4191 -----Original Message----- From: The IBM z/VM Operating System [mailto:[email protected]] On Behalf Of Daniel Tate Sent: Thursday, September 16, 2010 10:52 AM To: [email protected] Subject: Re: CP unresponsive on certain guests Output of Q SRM q srm IABIAS : INTENSITY=90%; DURATION=2 LDUBUF : Q1=100% Q2=75% Q3=60% STORBUF: Q1=300% Q2=250% Q3=200% DSPBUF : Q1=32767 Q2=32767 Q3=32767 DISPATCHING MINOR TIMESLICE = 5 MS MAXWSS : LIMIT=9999% ...... : PAGES=999999 XSTORE : 0% Ready; T=0.01/0.01 09:49:05 On Wed, Sep 15, 2010 at 5:47 PM, Dave Jones <[email protected]> wrote: > Hi, Daniel. > > The answer to your first question is to use the CP FORCE command (HELP > CP FORCE will tell you all about it.) The VM user id issuing the FORCE > command needs to have privilege class A as well. Usually this is done > from either MAINT or OPERATOR. > > The answer to your second question is a bit more difficult, I'm afraid. > As Marcy has already suggested, what does a Q SRM command show? My first > guess would be that your SLES11 guest is falling into Q3 and never given > an opportunity to run. > > To find out *why* the guest is not able to run, you need the services of > a good z/VM performance monitor.....IBM offers the Performance Monitor > (it comes bundles with z/VM, but it's an extra cost offering) and > Velocity Software (http://www.velocity-software.com/) has a very good > suite of products as well. IMHO it' practically impossible to run a > modern production grade z/VM-zLinux system without a good performance > monitor to help solve issues like the one your having now. > > On 09/15/2010 05:14 PM, Daniel Tate wrote: >> We're starting to run apps on the servers now. From time to time a >> guest will become unresponsive - to be more precise, ,the CP will not >> respond to commands, and neither will the guest OS (SLES11). not >> even #CP LOGOFF is acknowledged. from another login, CP INDIIC LOAD >> shows no appreciable load. >> >> Two questions from this: >> >> 1) how would I force a logoff of a user from another user? Is this possible? >> 2) if we are not paging and the IFLs are not loaded (2-3% utilization >> as a matter of fact) what could the bottleneck be? >> > > -- > Dave Jones > V/Soft Software > www.vsoft-software.com > Houston, TX > 281.578.7544 >
