Re: Is our cache_swap enough? studying store.log

map@iphil.net writes:

>Hi,
>
>I was thinking of a way of determining whether cache_swap is enough, or
>it needs to be increased, and more disks added.
>
>I was considering this:
>
>* If the average age of store.log RELEASE entries is less than the
> refresh_pattern specified, then the object is purged because we've
> ran out of space.

This doesn't seem quite right. For one thing, its hard to determine
a single age value for one refresh rule (you have to consider the
last modification times, etc), and even harder to determine an age
for a whole set of such rules.

Also, there are really only two reasons than an object is RELEASED.
Either it was bumped out for space, or a newer version was loaded
and the old was thrown out.

>* If it's equal, or greater, then Squid is just honoring the refresh_patterns
> and we have enough space.
>
>Any thoughts? It seems to me that determining cache_swap is still an art.
>We would like to add more space but not be prone to diminishing returns.

I think you might want to look at the 'LRU Age' value from the
cachemgr 'info' page

        Storage LRU Expiration Age: 10.41 days

Although this number can vary a bit from time to time, it roughly
tells you how often your entire cache is flushed.

There is a good paper at
    http://www.cs.ubc.ca/spider/marwood/Projects/SPA/Report/Report.html
which relates cache sizes and hit rates.

Duane W.

--
wessels@nlanr.net                           Think Globally, Cache Locally.