Re: Concerned about page fault ratio

On Wed, 15 Mar 2000, Steve Snyder wrote:

> I'm concerned about the page fault ratio (pfr) that I'm seeing (Squid
> 2.2S5, RedHat v6.0 /w kernel 2.2.14).
>
> The full output of "client mgr:info" is below, but here's the short
> version:
>
> pfr = Page_faults_with_physical_i/o / Number_of_HTTP_requests_received
>
> 0.47 = 9112 / 19368
>
> (The fault and requests numbers above are low due to my having
> rebooted recently, but the ratio calculated is representative of what
> I've been seeing.)
>
> This ratio is much higher than the examples I see in the Squid FAQ. I
> don't any reason for Squid to have to be swapped out. I've got plenty
> of RAM for the programs running on this box and the swap partition is
> barely used.
>
> Should I be concerned about the page fault ratio or do these numbers
> seem OK?

The most important thing is to watch the service time numbers.
If they get bad, or your cache seems really slow, then page fauls
may be the cause.

If response times seem good, then don't worry too much about
page faults.
Received on Wed Mar 15 2000 - 19:20:17 MST