Date: Tue, 21 May 1996 00:08:21 -0400 (EDT) From: Mark Mayo <mark@quickweb.com> To: Dave Andersen <angio@aros.net> Cc: freebsd-hackers@freebsd.org Subject: Re: Netscape Proxy Server on FreeBSD Message-ID: <Pine.BSF.3.91.960520235129.9352A-100000@scooter.quickweb.com> In-Reply-To: <199605202208.QAA13512@shell.aros.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 20 May 1996, Dave Andersen wrote: > Lo and behold, Mark Mayo once said: > > > I would be nice to see Squid packaged too! It's basically the > > continuation of the Harvest work (the non-commerical split..) and it > > works quite nicely! We have it setup in a hierarchical system, with a > > total of about 2 Gigs of cache. Works very well. So far it seems to be > > about 45% effective on cache hits. > > *whistle* How many customers do you have using the proxy? I'm beta > testing it with about 120 megs of cache and about .. I think 20-30 of my > dialin users, and getting a 16% non-local cache hit rate so far. I'd die > for 45% in the production cache, but it's probably a function of the > number of users. Actually, Squid is used in a University environment -- each department is behind an IP firewall, with a Squid proxy server on each gateway. Most departments use between 100MB to 500MB for their caches.. all of the departmental Squids are setup with their 'parent' being a central HP/UX machie on the pipe out of the U (the parent server has 2GB of cache space). The setup works superbly, because each department usually is interested in the same sort of sites, and sites that aren't in the local neighbor's disk are sent to the main server.. all in all we estimate about 45% success on cache hits. We were quite pleased. When we were beta testing on a single server (for the CIS department) we were getting about the same figures as you.. about 20% (again, the similar interests of the CIS dept. equals greater cache success!). Lesson of the day: squid rocks in an academic setting! In another setting (ISP), I'm running a Netsape Proxy (v1.12) -- but because it lacks the ideal architecture of the University's network, we only peak at 30% non-local cache hit success. That's with 1 GB and about 150-175 simultaneous users. I suspect you will see your percentages rise as you add more users. Again, I don't have any figures for the commercial setting, but I found that as more users came on, the more effective the proxy cache became. Be warned though - with 175 users, we often see our P100 with 48MB of RAM essentially fold under. Users hate it when the proxy doesn't respond...... under high loads, the best thing to do (at least with Netscape's product) is to give the cache all the breathing room it wants (at least 850 MB, probably more like 2GB) and only run the garbage collector late at night, not continusouly as the NS-proxy defaults too (just make sure you don't run out of disk space ;) For us, it was cheaper to throw a few extra gigs at the proxy and do the garbage-collection at non-peak times than it was to add more RAM and a faster processor and let the garbage-collector clean up expired docs all the time... Disk is the cheapest component in other words :-) Good luck with the caching! It can save you lots 'O dollars in the long run! -Mark :%t$sig -- Oops, thought I was in vi.. ------------------------------------------- | Mark Mayo mark@quickweb.com | | C-Soft www.quickweb.com | -------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.960520235129.9352A-100000>