[squid-users] How do I keep objects in memory in http accelerator mode??

   We have squid ( 2.5 STABLE 1 ) running as an http accelerator on a 2 processor SUN e420 with 2GB of memory. The site that it is accelerating have a small number of objects ( total size of all objects < 100 MB ) and each object ( a gif or jpg image ) is set to expire in 4 hours using apache mod_expires.

   What I'm confused about is that when squid is running, I see the process only taking up about 20MB and I notice that DISKD is active storing/retrieving from disk. Given the total number of objects and their total size, my preference would be to have squid avoid going to disk entirely and store all objects in memory.

   I tried setting the maximum_object_size to 0 and minimum_object_size to 500 KB. The comments in the squid.conf file suggest that these parameters would keep the objects from being stored to disk and I would assume that the objects would be cached in memory ( our cache_mem parameter is set to 300 MB with 2 cache directories of 200 MB each ). It seems like this should work, but when I try it all requests get passed back through to apache ( meaning, I don't think it's caching ).

   My question is, is it possible to force squid to cache only to memory ( or as much as possible ) and prefer going back to the server instead of disk?

   Thanks

    Andy
Received on Thu Oct 03 2002 - 12:32:43 MDT