Re: Large cache - performance tips

Adam Neat wrote:
>
> Hello,
>
> we appear to having problems with one of our clustered proxy servers.
>
> We're running a stable 1.1.22 on a secondary proxy server but is possibly
> having some peformance issues.
>
> We're running 106Gb of cache and 1Gb of RAM on this server.
>
> Its running on an Intel P II 450 box and hasnt had any problems in the
> past.
>
> A few clients have been suggesting that there are some performance issues,
> but then agian, we're getting nearly 20 - 25 million hits a day (when the
> winds blowing the right direction and its not raining) on a good day.
>
> On the local networks aorund the admin offices or operations centre, we can
> download files and pages with no problems or obvious performance issue.
>
> We run wiuth 64Mb of Ram as our cache_mem option and have the cache split
> over multiple SCSI II UW drives (9.1Gb UW IBM).
>
> the OS is a Red Hat 5.1 box with the normal security patches and the likes.
>
> any one have any tips or similar setup which may be providing good results?
>
> We dont use the swap so its not thrashing and the cpu doesnt reach above
> 0.20.
>
> Regards
>
> adam

Squid 1.1 serializes all disk requests.
Therefore a somewhat 'natural' upper limit is imposed on the number of
requests
the I/O subsystem can handle.
Suppose you've got the load spread across 10 disks providing an average
seek time of
9 ms plus average transaction time of 1 ms. this upper limit is
10 * 1/(0.009 + 0.001) = 1000 requests/sec.
These figures are just an example (overhead not accounted), but I guess
you get the idea.

My suggestion would be to go Squid 2, now that is has proven to be
stable.
It uses threads for disk I/O and thus allows the OS to cluster write
requests, greatly
decreasing the ratio seek vs. transaction time.

Markus