Re: Again, a performance problem on Digital Unix (OSF1)

In article <199703101344.OAA29592@josephine.sis.pasteur.fr> you write:

>After a long time of satisfying work, our cache has now started to stop
>again.
>
><h1>The symptoms</h1>
>
>We currently see 30 to 60 seconds of response time for retrieving a small
>HTML page already in the cache (measured by echoping or mconnect and
>logged in access.log). TCP connections to the cache manager are also
>incredibly slow.
>
>This occurs only when the system is loaded. During the week-end,
>everything is fine.
>
>The system worked fine a few days ago. We added three new cache children
>and performance suddenly dropped.

I have a Squid proxy with exactly the same symptoms and I posted some
experiences on the squid-dev lists for comments. As far as I can see there
is a strong relation between the number of parallel connections and the
response times. At some number of connections the response times are
getting up very fast which makes the number of connections go up even
further which makes the response times go up even more etc.

>2300 file descriptors.

That's an enormous amount of fd's, that must be approx. 1100 parallel
client connections.

Arjan
Received on Tue Mar 18 1997 - 00:06:36 MST