That depends. Linux kernel 2.4.13 may or may not have a badly broken VM
layer--which variant are you using? Linus or AC? If Linus, it probably
still had some bugs that might hit Squid, while AC definitely has the
old VM code that blows up with big processes. I ran into this at the
cacheoff quite badly.
The other dependency is, did you change the cache_mem settings between
the 1GB and 2GB Squid configurations? If the 2GB squid was configured
with a larger cache_mem, you might be hitting the problem of memory
fragmentation (no matter which kernel version). Check to see if you're
going into to swap at all on the 2GB configuration.
Also, going from 1GB to 2GB changes the memory addressing model of Linux
(and jumping from 2GB to 4GB does it yet again), but it shouldn't
severely impact performance--the added memory should more than make up
for it, in a workload like Squid.
Just some thoughts...might point you in the right direction.
Larry Creech wrote:
> Running squid on a dual 1ghz Athlon, I'm seeing squid top out about 475
> requests/sec on a 1 gigabyte box, and 425 requests/sec on the same box with
> 2 gig's of ram under Linux kernel version 2.4.13. Would this be a kernel or
> squid memory managment problem?
>
> -Larry
-- Joe Cooper <joe@swelltech.com> http://www.swelltech.com Web Caching Appliances and SupportReceived on Wed Dec 05 2001 - 11:39:33 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:05:14 MST