Re: [squid-users] Hardware Recommendations

-----BEGIN PGP SIGNED MESSAGE-----

Yeh I might be inclined to agree with raid being a bad
idea. The cache_dir method is quicker in my experience.
Unless you plan on running some form of striping over
the drives (which I'd say is slower anyway); and if your
thinking of redundancy; euw. Waste of drive space and
of speed.

Why not just have 10 cache_dir specs; and run with either
least-load or round-robin writing. Its squid using the drives;
allow it to decide what to do with them. Using RAID
will just slow you down.

However, as Matt Kehler has just mentioned, you really are
commiting a bit of an overkill here. I have spoken to people
with lot of experience about this; and from mine I agree
with the thought: there is no need for a fast box.

You ARE overspecing the box. 512M; 2 IDE drives, 1GIG
proc. should run fine to tell you the truth; and 2 IDE drives
is pushing it. You should remember that 200 users are not
200 new connections per second. They will not all be browsing
at once - and you need to allow for reading time. The load will
be split to about 40 at a time I'd say. (if they really are good
browsers)

As Joe Cooper and I have debated about before; a standard
box (on most *nix's) with 1 drive; 512meg of ram, and any
processor can run stable at about 50 - 60con's per sec; and
can max a good bit higher than that.

However, you know your users better than I; so to decide
about the RAID; I would NOT advise it. Go for multiple
cache_dir's

Dave.

[.......]

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 7.0.3 for non-commercial use <http://www.pgp.com>

iQEVAwUBPY1zT7TAuMekS5e2AQHHEQf/TRCrjEgzAYD5nCrwAAsm6+gdTQfQeSZD
WK2Wjh2XC5CGPjhVQhmSQTTl/60P5khW0K5fBT+G6IqMiNuEVmz+agaq6b5vv0qV
9TLKDXTGvIjiMoFuBFK/5jDFQy1XzFhMsGabGeZoquOGwcsVMwQPODM1jx4X+clh
KN+sZusyRrI0AxuYPvKXBYo7GnVrkYsJ3Bv5ITDY7V56zDi7z6oNy/mAC/sZVjjE
vDwUrr2MW1zf4hPqhB9kouPE5/9m0CUPRnwnwxbNrLYKaETBwHjRP76CvfxdxXVY
yFJTRYKwmYMsBOytgD0VhIIhLk6GPgda1D+EAVmKDTGMhN5/RenBUw==
=R2Ww
-----END PGP SIGNATURE-----
Received on Sun Sep 22 2002 - 01:42:59 MDT