Please help me on my squid 2.5 setup

[wennie]

Hello all,

Im having a problem with my squid setup, I have a server HP DL 380 3.06 Gig processor with 1 Gig memory and 3x64.6 scsi ultra 320 configure as raid 5. Im running named, Iptables and squid 2.5 stable on a Fedora Core 1.

My problem is that when there are few subscribers say 5 to 10 the squid seem to run smooth and fast but when i reach around 40 subscibers then it seems its very slow. below are the configuration i change in squid the rest are default:

http_port 3128
icp_port 3130
cache_mem 450 MB
cache_swap_low 90
cache_swap_high 95
cache_dir ufs /var/spool/squid 35000 256 512
cache_access_log /var/log/squid/access.log
cache_log /var/log/squid/cache.log
cache_store_log /var/log/squid/store.log
debug_options ALL,1
ftp_user Squid
acl ipaddr src 212.119.85.0/255.255.255.0
http_access allow ipaddr
cache_peer proxy.cyberia.net.sa parent 8080 3130 no-query
cache_peer 212.119.85.85 sibling 3128 3130
never_direct allow ipaddr
always_direct deny ipaddr


my second question, is download speed from squid to a subscriber is affected by the type of the program it access, example, php, jsp, html etc.

thank you very much and im hoping that you can help me on fine tuning my squid server.

wennie

[UnixMan]

Very sorry that I can't help with this - I've never used the product. Can anyone else reading this help?

[kjteoh]

not so much in the /etc/squid/squid.conf but lets consider other things.
Filesystem/RAID - you can imagine that lots of i/o will happen and if you are running a squid server big time, you should be using raid0

the ext2/3 filesystem is rather slow .. reiserfs i experience is better for this kind of stuff.

next - iptables? are there any rate limits?

last, run bind as cacheing nameserver to speed up dns request. you may also need to enable ident 113 at firewall level.

ive run squid with 200+ abusive users on half the machine you have w/o complaints. further i included authentication and spawn 12 ncsa authenticators.

also look closely at /var/log/squid/access.log and run calamaris over it to see if there is a bottleneck by some user. if so, you will need to configure tc/iproute2 and cbq/htb to control bandwidth.