Date: Thu, 21 Oct 2010 15:42:37 -0700 From: David Wolfskill <david@catwhisker.org> To: performance@freebsd.org Subject: Re: Possible evidence of performance regression for 8.1-S (vs. 7.1) Message-ID: <20101021224237.GG52404@albert.catwhisker.org> In-Reply-To: <20101021215330.GA86224@dan.emsphone.com> References: <20101020174854.GZ21226@albert.catwhisker.org> <20101021215330.GA86224@dan.emsphone.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--SdvjNjn6lL3tIsv0 Content-Type: multipart/mixed; boundary="w/VI3ydZO+RcZ3Ux" Content-Disposition: inline --w/VI3ydZO+RcZ3Ux Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Thu, Oct 21, 2010 at 04:53:30PM -0500, Dan Nelson wrote: > ... > An observation: on 8.1, both user and sys times are less, but real time = is > higher. So 8.1 finished the build using less CPU, but spent more time > waiting for something else. Interesting; thanks for pointing that out. > Disk? Network? I don't suppose the machines > are low enough on RAM that you end up swapping at any point? They have 6GB RAM each, but we're running FreeBSD/i386 without PAE. Still, the workloads under test were the only things running (other than fairly constant administrivia) on each of them, and the build is not especially memory-constrained, as far I as I know. As far as disk, during the (vast) majority of the execution time, disk I/O occurs at such a low rate that it's barely noticeable. And these were local builds -- that is, done on local disk (vs. NFS). The captive build tools were accessed via NFS, but that part is the same on each machine. I've attached a file that shows some of the additional rusage fields from /usr/bin/time -- it's fairly wide (240 chars) -- that may provide some additional insight. Some of the differences I note: * maxrss is almost identical (but slightly higher for 8.1 vs. 7.1). * avgsms for 8.1 is small in each case, but 8.1 is about 20 - 30% of 7.1. * pgrec is a bit lower in 8.1. * pf varies a bit, but looks about the same for each. * sw is 0 for all runs (which is as expected). * bio is noticeably lower in 8.1. * boo is also noticeably lower in 8.1. * msgs is also noticably lower in 8.1. * msgr is just about identical in all cases. * sigr is lower in 8.1. * 8.1's vcsw is about 60% of 7.1's. * 8.1's icsw is about 50% of 7.1's. [Each of the above percentages is an "eyeball estimate," so use an appropriate grain of salt.] > Maybe there > was a change to /usr/bin/make that is causing it to launch jobs slower?= =20 We use "captive" versions of make, so the version on the system is a moot point. > Julian's suggestion of booting the 8.1 kernel on the 7.1 OS will definite= ly > narrow down the list of suspects. I'll see about doing something along those lines, but I doubt it will be all that helpful, actually. --=20 David H. Wolfskill david@catwhisker.org Depriving a girl or boy of an opportunity for education is evil. See http://www.catwhisker.org/~david/publickey.gpg for my public key. --w/VI3ydZO+RcZ3Ux Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename=time_stats start stop real user sys maxrss avgsms pgrec pf sw bio boo msgs msgr sigr vcsw icsw os 1287436357 1287461948 25590.99 81502.22 18115.07 403408 40 1632501960 70091 0 1526450 3948957 52333862 16776 414554 116146505 69359713 FreeBSD 8.1-STABLE i386 1287462797 1287488766 25969.26 81452.14 17920.14 403416 33 1632275106 70921 0 1569112 3992055 52289209 16780 415153 112395968 63795293 FreeBSD 8.1-STABLE i386 1287489641 1287515287 25645.84 81548.40 18256.52 403408 44 1633037428 68757 0 1623321 3939857 52405680 16782 415519 120381724 72994804 FreeBSD 8.1-STABLE i386 1287516151 1287541481 25329.64 81546.23 18294.10 403416 44 1632579746 67606 0 1510313 3932011 52341946 16776 415086 120137651 74621868 FreeBSD 8.1-STABLE i386 1287542355 1287568599 26244.59 81431.47 17902.39 403416 30 1632203678 68433 0 1522405 3959469 52304716 16776 415022 111403586 62745487 FreeBSD 8.1-STABLE i386 1287525363 1287546846 21483.13 82628.20 21703.09 403344 170 1646120708 69576 0 1714249 4096652 66054010 16776 426045 181765163 138243214 FreeBSD 7.1-RELEASE-JNPR-4 i386 1287548005 1287569100 21094.63 82853.19 22185.02 403344 194 1646134922 69189 0 1731574 4130719 66054759 16776 425545 200096792 162326694 FreeBSD 7.1-RELEASE-JNPR-4 i386 1287570300 1287591371 21071.33 82756.81 21943.22 403348 184 1646103669 70720 0 1841404 4097966 66064504 16786 426741 187956827 152273232 FreeBSD 7.1-RELEASE-JNPR-4 i386 1287592592 1287614103 21511.23 82637.30 21849.90 403344 177 1646110915 69878 0 1704692 4088578 66057042 16776 424758 189088692 145093952 FreeBSD 7.1-RELEASE-JNPR-4 i386 1287615323 1287636770 21446.42 82715.81 21708.97 403344 169 1646267475 68996 0 1702403 4096640 66087016 16776 426374 175517603 138999711 FreeBSD 7.1-RELEASE-JNPR-4 i386 --w/VI3ydZO+RcZ3Ux-- --SdvjNjn6lL3tIsv0 Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.16 (FreeBSD) iEYEARECAAYFAkzAwdwACgkQmprOCmdXAD0rGwCfaB9t5X3cbjbs8QJyDMsxt20V BvwAn2m6HZ/HlPkHyureCvvNEItwCWGZ =i9MB -----END PGP SIGNATURE----- --SdvjNjn6lL3tIsv0--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20101021224237.GG52404>