Jeff Peng wrote:
Hello list,
i'm interested in some numbers other people have when using squid as a
cache proxy.
Our setup will be 5000 concurrent users surfing on the web.
I have a dl380 g5 with 8 10k rpm disks and 32GB ram
Sounds a good hardware box.
Would that be sufficient or should i run two squid's on that box (using
more cpu's)
Or should i use two boxes.
If you run two squid instances on a server box, they have to listen on
different TCP ports. If you don't mind that case, it's fine.
I would assign two ip adresses to that box for running two squid's
are 5000 concurrent users a lot are do other people have more users
How much concurrent connections they can make?
My squid boxes always got 20,000+ concurrent connections.
Each server box has 4G mem, with a fast SCSI disk.
i would assigh 12GB memcache and four disks with diskd to each instance.
What (linux) os kernel settings should i use?
ulimit stuff and and what scheduler?
Which version off squid should i use (2.7 3.0 3.1)
Both are fine.
But people here said squid-2.7 has better performance for high traffic.
Regards,
Jeff.
Thanks for the quick reply
