From what little I understand, async-io should help squid access the cache
faster because multiple threads will be accessing the cache concurrently.
According to the Squid doc (in the squid.conf.default fle), the async-io
support might be a little buggy so be careful.
I've been running a Squid-2.3.STABLE1 proxy on a Pentium 166 /w a 200MB
(now 500MB) cache serving between 15 - 40 clients with async-io for a few
weeks now without problems.
You can always control this with the cache_dir configuration setting.
Build Squid with --enable-async-io on and specify the cache_dir like this:
cache_dir /usr/local/squid/cache asyncfs 100 6 256
If you want to stop using async-io, then change the "asyncfs" to "ufs" and
reconfigure Squid (squid -k reconfigure). The "asyncfs" parm might
actually be "asyncufs", but the exact syntax will be in the
squid.conf.default file.
Good luck!
Peace.....
Tom
"Jason Thompson" <jthompson@connect4free.net> on 03/10/2000 01:33:09 PM
To: squid-users@ircache.net
cc: (bcc: Tom Williams/HQ/dssi)
Subject: async-io or no aysnc-io
Hi All,
I will be setting up a squid server soon, and was hoping if someone could
advise me if I should use async-io or not. Please could you explain to me
the reason for having this option.
The server spec will be:
Intel Penitum 166
256Mb RAM
2 x 2GB IDE Hard Disks
1 x 100Mb Network Card
OS: Redhat 6.0
Internet Connection: 64kbs (soon to be 128kbs)
This spec will be serving a maximum of 100 clients, in a school. Will this
spec be enough?
Finally, anyone know if Corel Linux Deluxe is (a) any good, and (b)
suitable for running squid?
Thanks
Jason
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:52:03 MST