Re: [squid-users] Squid LAN transfer performance slow

From: Marc Elsen <marc.elsen@dont-contact.us>
Date: Sun, 27 Oct 2002 17:08:02 +0100

"Tais M. Hansen" wrote:
>
> On Sunday 27 October 2002 16:20, Marc Elsen wrote:
> > Using 'top' can you SQUID's mem usage ?
> > Especially compare SIZE versus RSS,is there any significant difference
>
> Squids primary process SIZE and RSS is exactly 1148. There's only little
> difference between SIZE and RSS on all children. SIZE is 5804 and RSS is
> 5784.
>
> > Check SQUID's cache.log for possible error messages during
> > normal operation.
>
> Nothing stands out.
>
> > Check 'dmesg' on Linux ; watch for suspicous error messages,if any.
> > Check /var/log/messages on Linux, idem ditto.
>
> None either.
>
> > Check 'netstat -i' ; watch for error counters.
>
> Watched it for about a minute while stressing squid. 0 errors on all ifs in
> both directions.
>
> > Check 'netstat -a' , TCP connections should not be in suspicous
> > states.
>
> Seems okay. Anything in particulair?
>
> Just tried requesting a large file (30MB) through squid on localhost. It also
> only transfers about 200-250 KB/s. Tried increasing squids priority to max
> without any significant change. It doesn't look like the harddisk could be
> the bottleneck since there's almost no HD activity while xfering.

 Could there be induced bottlenecks as a result of the lan perimeter
 parameters ?
 What kind of connection is being realized to the Intranet.
 Check port status on network hub, watch for possible errors over there.

 If the same file size is being transferred from the SQUID machine
 to a SQUID client, without SQUID, I mean , just ftp a file
 from the machine to the Intranet, are then the same bottlenecks
observed ?

 M.

>
> --
> Regards,
> Tais M. Hansen
> OSD

-- 
 'Time is a consequence of Matter thus
 General Relativity is a direct consequence of QM
 (M.E. Mar 2002)
Received on Sun Oct 27 2002 - 09:08:05 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:10:55 MST