From owner-pcp@oss.sgi.com Thu Apr 19 08:54:41 2001 Received: (from majordomo@localhost) by oss.sgi.com (8.11.3/8.11.3) id f3JFsfW15037 for pcp-outgoing; Thu, 19 Apr 2001 08:54:41 -0700 Received: from alsace.it-innovation.soton.ac.uk (glass.it-innovation.soton.ac.uk [152.78.239.150]) by oss.sgi.com (8.11.3/8.11.3) with ESMTP id f3JFseM15034 for ; Thu, 19 Apr 2001 08:54:40 -0700 Received: by ALSACE with Internet Mail Service (5.5.2650.21) id <2LBL9CTF>; Thu, 19 Apr 2001 16:52:23 +0100 Message-ID: <988586BC1026D511B55B0060087E5D4705728A@ALSACE> From: Mike Jewell To: "'pcp@oss.sgi.com'" Subject: PCP and processes Date: Thu, 19 Apr 2001 16:52:23 +0100 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2650.21) Content-Type: text/plain; charset="iso-8859-1" Sender: owner-pcp@oss.sgi.com Precedence: bulk Hi, I want to do some process benchmarking (getting the memory usage/disk usage/CPU usage/etc while a program is running) - can I use PCP to handle this? I've seen lots of examples which monitor system performance, but nothing at a process level? Cheers, Mike Jewell From owner-pcp@oss.sgi.com Sun Apr 22 19:19:01 2001 Received: (from majordomo@localhost) by oss.sgi.com (8.11.3/8.11.3) id f3N2J1S32339 for pcp-outgoing; Sun, 22 Apr 2001 19:19:01 -0700 Received: from yog-sothoth.sgi.com (eugate.sgi.com [192.48.160.10]) by oss.sgi.com (8.11.3/8.11.3) with ESMTP id f3N2IpM32328 for ; Sun, 22 Apr 2001 19:18:51 -0700 Received: from rattle.melbourne.sgi.com (rattle.melbourne.sgi.com [134.14.55.145]) by yog-sothoth.sgi.com (980305.SGI.8.8.8-aspam-6.2/980304.SGI-aspam-europe) via ESMTP id EAA8231858 for ; Mon, 23 Apr 2001 04:18:48 +0200 (CEST) mail_from (kenmcd@melbourne.sgi.com) Received: from localhost (kenmcd@localhost) by rattle.melbourne.sgi.com (SGI-8.9.3/8.9.3) with ESMTP id MAA82865; Mon, 23 Apr 2001 12:16:35 +1000 (AEST) X-Authentication-Warning: rattle.melbourne.sgi.com: kenmcd owned process doing -bs Date: Mon, 23 Apr 2001 12:16:35 +1000 From: Ken McDonell Reply-To: To: Mike Jewell cc: "'pcp@oss.sgi.com'" , Mark Goodwin Subject: Re: PCP and processes In-Reply-To: <988586BC1026D511B55B0060087E5D4705728A@ALSACE> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-pcp@oss.sgi.com Precedence: bulk On Thu, 19 Apr 2001, Mike Jewell wrote: > Hi, > > I want to do some process benchmarking (getting the memory usage/disk > usage/CPU usage/etc while a program is running) - can I use PCP to handle > this? I've seen lots of examples which monitor system performance, but > nothing at a process level? On IRIX we have a large number of per-process metrics to track resource demands. On Linux, things are considerably more spartan. You may finds something of use in the following subset (all of which can be monitored with PCP for some or all processes). Mark, I think the "this and all it's child processes" description is not correct ... the c variants appear to be the cummulative resource consumption for all children that have previously exited, so this process and still running children are _not_ included. bruce 2% pminfo -T proc ... proc.psinfo.minflt Help: count of minor page faults (i.e. reclaims) proc.psinfo.cmin_flt Help: count of minor page faults (i.e. reclaims) for the process and its children proc.psinfo.maj_flt Help: count of page faults other than reclaims proc.psinfo.cmaj_flt Help: count of page faults other than reclaims for the process and its children proc.psinfo.utime Help: time (in ms) spent executing user code since process started proc.psinfo.stime Help: time (in ms) spent executing system code (calls) since process started proc.psinfo.cutime Help: time (in ms) spent executing user code for this and all it's child processes proc.psinfo.cstime Help: time (in ms) spent executing system code for this and all it's child processes ... proc.psinfo.vsize Help: virtual size of the process in Kbytes proc.psinfo.rss Help: resident set size (i.e. physical memory) of the process ... proc.psinfo.nswap Help: count of page swap operations proc.psinfo.cnswap Help: count of page swap operations for this process and all it's child processes ... proc.memory.size Help: instantaneous virtual size of process, excluding page table and task structure. proc.memory.rss Help: instantaneous resident size of process, excluding page table and task structure. proc.memory.share Help: instantaneous amount of memory shared by this process with other processes proc.memory.textrss Help: instantaneous resident size of process code segment in Kbytes proc.memory.librss Help: instantaneous resident size of library code mapped by the process, in Kbytes proc.memory.datrss Help: instantaneous resident size of process data segment, in Kbytes proc.memory.dirty Help: instantaneous amount of memory that has been modified by the process, in Kbytes From owner-pcp@oss.sgi.com Sun Apr 22 20:17:54 2001 Received: (from majordomo@localhost) by oss.sgi.com (8.11.3/8.11.3) id f3N3Hs102176 for pcp-outgoing; Sun, 22 Apr 2001 20:17:54 -0700 Received: from yog-sothoth.sgi.com (eugate.sgi.com [192.48.160.10]) by oss.sgi.com (8.11.3/8.11.3) with ESMTP id f3N3HnM02172 for ; Sun, 22 Apr 2001 20:17:49 -0700 Received: from larry.melbourne.sgi.com (larry.melbourne.sgi.com [134.14.52.130]) by yog-sothoth.sgi.com (980305.SGI.8.8.8-aspam-6.2/980304.SGI-aspam-europe) via SMTP id FAA8245378 for ; Mon, 23 Apr 2001 05:17:46 +0200 (CEST) mail_from (markgw@sgi.com) Received: from sandpit.melbourne.sgi.com (sandpit.melbourne.sgi.com [134.14.55.132]) by larry.melbourne.sgi.com (950413.SGI.8.6.12/950213.SGI.AUTOCF) via ESMTP id NAA04159; Mon, 23 Apr 2001 13:15:32 +1000 Date: Mon, 23 Apr 2001 13:15:31 +1000 (EST) From: Mark Goodwin X-Sender: markgw@sandpit.melbourne.sgi.com To: kenmcd@sgi.com cc: Mike Jewell , "'pcp@oss.sgi.com'" Subject: Re: PCP and processes In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-pcp@oss.sgi.com Precedence: bulk On Mon, 23 Apr 2001, Ken McDonell wrote: > Mark, I think the "this and all it's child processes" description is > not correct ... the c variants appear to be the cummulative > resource consumption for all children that have previously exited, > so this process and still running children are _not_ included. The c variants do not include the process itself, nor any children that are still running or defunct/zombie. i.e. the value is updated when a child exits and the parent waits. Not unexpectantly, the init process and most shell processes have relatively large values for these metrics ... Help text now reads: proc.psinfo.min_flt count of minor page faults (i.e. reclaims) proc.psinfo.cmin_flt count of minor page faults (i.e. reclaims) for all children of the process that have exited and were waited on by the process proc.psinfo.maj_flt count of page faults other than reclaims proc.psinfo.cmaj_flt count of page faults other than reclaims for all children of the process that have exited and were waited on by the process proc.psinfo.utime time (in ms) spent executing user code since process started proc.psinfo.stime time (in ms) spent executing system code (calls) since process started proc.psinfo.cutime time (in ms) spent executing user code for all children of the process that have exited and were waited on by the process proc.psinfo.cstime time (in ms) spent executing system code for all children of the process that have exited and were waited on by the process