Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 7 Oct 1999 15:04:53 -0300 (ADT)
From:      "Marc G. Fournier" <marc.fournier@acadiau.ca>
To:        squid-users@ircache.net
Cc:        freebsd-hackers@freebsd.org
Subject:   Squid under FreeBSD 3.3-STABLE ...
Message-ID:  <Pine.BSF.4.10.9910071445530.543-100000@atelier.acadiau.ca>

next in thread | raw e-mail | index | archive | help

I'm running it right now on a 3.3-STABLE machien with a 166Mhz CPU (there
are two inthe machine, but FreeBSD won't allow me to bring up the second
oen *sigh*)...the machine has 512MB of RAM and ~24Gig of Disk space...

The server is configured to use 128MB of RAM, and top is currently 
reporting:

39237 root       2   0   201M   200M poll   210:10 11.33% 11.33% squid

And swap space isn't even being touched:

demeter# pstat -s
Device      1K-blocks     Used    Avail Capacity  Type
/dev/da0s1b   1053584        0  1053456     0%    Interleaved
/dev/da1s1b    262144        0   262016     0%    Interleaved
/dev/da2s1b    262144        0   262016     0%    Interleaved
/dev/da3s1b    262144        0   262016     0%    Interleaved
/dev/da4s1b    262144        0   262016     0%    Interleaved
/dev/da5s1b    262144        0   262016     0%    Interleaved
/dev/da6s1b    262144        0   262016     0%    Interleaved
Total         2625552        0  2625552     0%

All in all, I'm impressed...we did >490k requests through that server
yesterday, after only ~7days since we brought it online, with 1800 hosts
using it, and an average of 36% hit rate...

Looking at the PolyGraph/BakeOff report, it has "FreeBSD Tweaks" in a
section to its own...except that it pertains to FreeBSD 2.2.8, so I'm
wondering both how much of it still applies as well as whether there is
anything new I should be looking at...

It lists:

1. Increase per process memory limits (already done)

2. Increase limit on mbuf clusters
	- netstat -m doesn't seem to indicate it being required, but it
	  can't hurt
	- did it

3. Disabled TCP delayed acknowledgements (options TCP_ACK_HACK)...??
	- can't find a reference to this one...was it just a 2.2.8ism?
	- didn't do it 

4. Decreased TCP's Maximum Segment Lifetime to three seconds
	- did it

5. Apply a patch to uipc_socket.c to do with "fixing TCP bugs with small
   packets?
	- didn't do it

6.  Increased system wide and per proces FD limits
	- did it

7. Increased ephemeral port range to 30,000 (net.inet.ip.portrange.last)
	- did it

8. Insrease the socket listen queue size (kern.somaxconn)
	- doesn't appear to exist anymore

9. Set resources limits in /etc/login.conf
	- did it

10. Disable unnecessary processes
	- didn't do this, since am using them

This document can be found at http://bakeoff.ircache.net ...

Comments/ideas on how to improve things?

Thanks...

Marc G. Fournier                                 marc.fournier@acadiau.ca
Senior Systems Administrator                            Acadia University

  "These are my opinions, which are not necessarily shared by my employer"





To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.10.9910071445530.543-100000>