On 9/13/07, Ben Rockwood <[EMAIL PROTECTED]> wrote: > I've been using DTrace more and more to investigate storage performance > issues and keep bumping into something I can't properly explain. The > following snipped comes from Brenden's 'iotop': > > ... > 102 13171 11536 pickup sd6 31 384 R 2364416 > 80 14597 10632 httpd sd5 31 320 R 2763776 > 100 5089 4977 mysqld sd5 31 320 W 4349952 > 0 0 0 sched sd5 31 320 W 5414912 > 0 0 0 sched sd3 31 192 R 8247296 > 0 0 0 sched sd4 31 256 R 11098112 > > This is on a snv_67 system with several active zones. I've been ignoring > the 'sched' IO but I'd like to understand it. Can someone clue me in on why > I see so much IO that appears to actually be touching disk?
If I recall correctly, the writes are associated with dirty pages being flushed from the page cache, and the reads are typically associated with file system read ahead activity. I am basing this on something Brendan posted a long long time ago, but I can't seem to find his original message to verify this. Brendan -- if I'm wrong, please correct me. Thanks, - Ryan -- UNIX Administrator http://prefetch.net _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
