On Sun, 7 Feb 1999, John Sloan wrote:
> Date: Sun, 7 Feb 1999 18:29:55 +0000 (GMT)
Sorry for not replying sooner; it's taken me this long to get back to reading
my squid-users inbox. :-(
> Ah yes. Async IO. My average CPU useage increased noticeably when I
> enabled it. For the request rate I'm seeing [peaks of 30/sec on average]
> I needed to configure NUMTHREADS to be 64 in aiops.c before it wouldn't
> moan about running out on a regular basis.
Combined HTTP/ICP requests to one of my proxies shows (cachemgr.cgi) around
49.5 requests per second (presumably that's an average over the period of
time Squid's been running - on that box, it's about a week at the moment).
I haven't modified NUMTHREADS at all (so presumably it's still just using 16).
Unfortunately the cache.log has rotated from yesterday (and autodeleted) so a
grep for "thread" doesn't show any error at the moment.
Median service times versus CPU usage:
Median Service Times (seconds) 5 min 60 min:
HTTP Requests (All): 0.55240 0.52331
Cache Misses: 0.76407 0.85130
Cache Hits: 0.03829 0.04519
Near Hits: 1.00114 0.94847
Not-Modified Replies: 0.02742 0.02899
DNS Lookups: 0.02447 0.01940
ICP Queries: 0.00647 0.00687
Resource usage for squid:
CPU Usage: 28.68%
CPU Usage, 5 minute avg: 19.80%
CPU Usage, 60 minute avg: 28.95%
The box is dual CPU (167MHz UltraSPARC-I, 512Kb L2 cache) running nothing but
Squid (and its associated processes).
Cheers..
dave
Received on Mon Feb 15 1999 - 08:22:40 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:44:33 MST