Thanks Ralf, > Which kernel?
Kernel is 2.4.20-SMP (directly from Redhat 9 ). > Squid is probably I/O bound. And due to it's architecture it cannot take advantage of another processor. I didn't know that. > RAID 1 doesn't help either. Better use different disks or a RAID 0 -- after all Squid's data partitions are expendable! > The money would have been better invested in more disks, not more processors. Plan before you buy! That's not so far of what I'm doing: I'm currently in the test period. Real servers will be different (more disks with different volumes, more powerful). I'm interested in basic performance test right now in order to now what I have to require for my real servers ;) > Partitions are extfs3. >Uh, that's bound to be slow. Mount the data area for Squid noatime, maybe in writeback mode, instead of ordered mode. That will be faster. > Still faster are XFS and ReiserFS. >We use this machine: > >Dell Poweredge (don't know which) >2GB RAM >Intel(R) Xeon(TM) CPU 2.80GHz >/dev/sda8 on / type auto (rw,errors=remount-ro) >proc on /proc type proc (rw) >devpts on /dev/pts type devpts (rw,gid=5,mode=620) >/dev/sdb5 on /squid-cache0 type xfs (rw) >/dev/sdc5 on /squid-cache1 type xfs (rw) >/dev/sdd5 on /squid-data type xfs (rw) >/dev/sda5 on /boot type ext3 (rw) >Filesystem Size Used Avail Use% Mounted on >/dev/sda8 13G 5.7G 6.5G 47% / >/dev/sdb5 17G 15G 2.5G 86% /squid-cache0 >/dev/sdc5 17G 15G 2.5G 86% /squid-cache1 >/dev/sdd5 17G 5.0G 12G 30% /squid-data >/dev/sda5 69M 9.0M 56M 14% /boot >At a real-world load (production use) of 200 connections/s it has a load of 0.75. You mean 0.75% of CPU Load ??? Impressive !!! Is it 200 connection/s = 200 request/sec ?? Thank you very much, Nicolas Chaillot