From: DAUBIGNE Sebastien - BOR ( SDaubigne@bordeaux-bersol.sema.slb.com ) (SDaubigne@bordeaux-bersol.sema.slb.com)
Date: Tue Jul 22 2003 - 12:49:22 EDT
We have a Solaris 2.6/Oracle box which has poor throughput and a high (from
50 to 100) number of IO busy processes (column "b" of vmstat).
CPU (50%)/memory (no paging) are OK, so I assume the poor throughput is due
to the disk part.
Now, I'm trying to identify the disks which are the source of the bottleneck
with "iostat", but I can hardly figure out a general rule (is there any ?)
to do this : what are the most significant column of "vmstat" to look at,
and the corresponding threshold ?
Various perf docs state that disk stats are "bad" if :
- %b > 65 ("disk too busy to handle the load correctly")
or
- asvc_t > 30 ("disk too slow")
Do you think theses columns/values are good candidates ?
As for asvc_t value, it depends on the block size, so that 60 ms is good for
a 1 Mb I/O (sequential) but bad for a 16 Kb one (random), right ? So, what
about this 30 ms threshold ? Should I link asvc_t with average I/O size ?
Also, what is the good interval for iostat samples : 30 sec ? 5 min ?
Any advice on this topic will be appreciated.
I will summarise.
--- Sebastien DAUBIGNE sdaubigne@bordeaux-bersol.sema.slb.com <mailto:sdaubigne@bordeaux-bersol.sema.slb.com> - (+33)5.57.26.56.36 SchlumbergerSema - SGS/DWH/Pessac _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:26:47 EDT