Re: [squid-users] huge page faults

Van Bossche Koen wrote:
>
> The machine it runs on is a Celeron 533 with 256Mb Ram, 20Gb IDE, 3Com
> ethercard
> OS is Linux Redhat v6.2
> Squid (2.5DEV-PRE4) is configured with 64MB memory / 4 GB cache
>
 ...

 Concerning the page faults : - you should using 'top' compare
 the SQUID process' mem usage parameters , especially SIZE column
 versus RSS column (RSS=resident set size).

 If the SIZE is 'much bigger' then RSS , then it can be assumed
 that SQUID will be swapping a lot, because it needs more memory
 then there is physically available.

 The RSS is the chunk of the process in physical memory.

 Perhaps with 256Gb it would be advisable to lower the
 cache_mem setting to 16Mb for instance, because I read
 somewhere that SQUID total process size can be 3 or 4 times
 this value.
 
 Or , the money based solution : if SIZE (of SQUID) is much more
 then RSS , then you may decide to add physical memory.

 M.
Received on Thu Apr 04 2002 - 10:09:24 MST