Re: [squid-users] Acceptable Service Times?

On Mon, 28 Jun 2010 15:09:53 -0500, "Baird, Josh" <jbaird_at_follett.com>
wrote:
> Hi,
>
> I have a pair of forward proxy's (Squid 2.6STABLE 21/EL5 averaging about
> 500-600requests per minute currently. All caching has been disabled.
> Some users are reporting high latency and slow browsing. Below is a
> snapshot of stats from Squid. Could someone tell me if anything stands
> out that would cause a problem? Service times look to be acceptable to
> me, but perhaps I am looking at them incorrectly?

As long as they are acceptable to you and your clients.

DNS could possibly be improved, but unless its causing issues or your
about to gain lots of clients its nothing to worry about.

>
> You can also view a graph of 5min service times at
> http://thunder.jbdesign.net/~jbaird/servicetimes.png
>
> Squid Object Cache: Version 2.6.STABLE21
> Start Time: Sun, 27 Jun 2010 15:17:41 GMT
> Current Time: Mon, 28 Jun 2010 20:04:09 GMT
> Connection information for squid:
> Number of clients accessing cache: 5
> Number of HTTP requests received: 848091
> Number of ICP messages received: 0
> Number of ICP messages sent: 0
> Number of queued ICP replies: 0
> Request failure ratio: 0.00
> Average HTTP requests per minute since start: 491.2
> Average ICP messages per minute since start: 0.0
> Select loop called: 20321720 times, 5.097 ms avg
> Cache information for squid:
> Request Hit Ratios: 5min: 0.0%, 60min: 0.0%
> Byte Hit Ratios: 5min: 0.7%, 60min: 1.0%
> Request Memory Hit Ratios: 5min: 0.0%, 60min: 0.0%
> Request Disk Hit Ratios: 5min: 0.0%, 60min: 0.0%
> Storage Swap size: 0 KB
> Storage Mem size: 288 KB
> Mean Object Size: 0.00 KB
> Requests given to unlinkd: 0
> Median Service Times (seconds) 5 min 60 min:
> HTTP Requests (All): 0.06286 0.05633
> Cache Misses: 0.06286 0.05633
> Cache Hits: 0.00000 0.00000
> Near Hits: 0.00000 0.00000
> Not-Modified Replies: 0.00000 0.00000
> DNS Lookups: 0.00573 0.00669
> ICP Queries: 0.00000 0.00000
<snip>

2.6 is still very IO driven, so AFAIK its service efficiency declines with
lower traffic (which possibly explains your 'worse' graphed times over the
weekend, in particular early am).

Amos
Received on Mon Jun 28 2010 - 22:47:01 MDT