Date: Thu, 9 Feb 2006 00:54:51 +0530 From: Joseph Koshy <joseph.koshy@gmail.com> To: astatech@mail.ru Cc: freebsd-smp@freebsd.org Subject: Re: 6.0 Stable on 8 Core AMD Opteron scalability. (perf stat) Message-ID: <84dead720602081124o651693ccy7920658a54d97e61@mail.gmail.com> In-Reply-To: <E1F6ohz-0001cf-00.astatech-mail-ru@f64.mail.ru> References: <84dead720602070831g32e14257x16841f31e5b8527d@mail.gmail.com> <E1F6ohz-0001cf-00.astatech-mail-ru@f64.mail.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
jk> which benchmark? a> jakarta jmeter with customly builded test plan for PHPBB No Java on my system :(. a> bge0: flags=3D8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 ... a> media: Ethernet autoselect (none) a> status: no carrier It wasn't clear if you were running over a localhost-only benchmark. If not, did your network card manage to negotiate full-duplex mode? a> See below... a> PS: maybe httpd wait for log write? a> $ vmstat 2 99999 ... [vmstat output] a> 66 0 0 1375484 14423592 34484 0 0 0 31638 0 15 0 3503 259224 109744 53 21 26 Ok, no memory pressure, a large system-call rate and a large number of context switches per second. a> 8634 mysql 111 76 0 367M 122M RUN 7 0:52 0.00% mysqld This process looks suspicious. Please check if you are running MySQL with an appropriate configuration (my-huge.cnf?). http://wikitest.freebsd.org/moin.cgi/MySQL has more hints about tuning MySQL. Does changing sysctl kern.timecounter help? I would appreciate a hwpmc profile of your system under this load. Step 1 # kldload hwpmc 2 # pmcstat -S k8-bu-cpu-clk-unhalted -O /tmp/logfile [for less than a minute] 3 ^C [kill pmcstat] 4 # pmcstat -R /tmp/logfile -g 5 # [for each foo.gmon file generated, run gprof -l /path/to/unstripped/foo foo.gmon to get a flat profile] -- FreeBSD Volunteer, http://people.freebsd.org/~jkoshy
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?84dead720602081124o651693ccy7920658a54d97e61>