Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.cs.su.oz.au!metro!metro!munnari.OZ.AU!news.mel.connect.com.au!news.syd.connect.com.au!news.bri.connect.com.au!fjholden.OntheNet.com.au!not-for-mail From: Tony Griffiths <tonyg@OntheNet.com.au> Newsgroups: comp.unix.bsd.freebsd.misc Subject: Re: Formating Drives For Squid Date: Tue, 08 Apr 1997 10:32:53 +1000 Organization: On the Net (ISP on the Gold Coast, Australia) Lines: 60 Message-ID: <33499235.6FF0@OntheNet.com.au> References: <3346eefa.0@news.topend.com.au> <5i6459$en2$1@klemm.gtn.com> Reply-To: tonyg@OntheNet.com.au NNTP-Posting-Host: swanee.nt.com.au Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0 (WinNT; I) To: nagy@darwin.topend.com.au Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:38753 Andreas Klemm wrote: > > In article <3346eefa.0@news.topend.com.au>, > nagy@darwin.topend.com.au (Robert Nagy) writes: > > Hi Folks, > > > > I'll be formatting a couple of drives to use as disk cache with squid. > > What is the best block size (& other settings) to set these drives up as. > > Both drives are 3.1GB in size. > > The new squid's cachemgr.cgi interface shows you how many > percent of objects use a certain block size ... I've done a bit of research in this area and have just installed a new cacheing proxy. Basically, the average HTTP object seems to be about 11-12 KB while the average FTP object is 2 orders of magnitude greater (ie. 0.5 - 1 MB) but there are FAR fewer of them. The ratio of HTTP to FTP objects in our cache is 230:1 ! > So I'd first suggest you use the standard ufs filesystem block- > and fragsize, 8k/1k, but increasing the number of inodes, > since I assume, you get similar little files as in a news > filesystem. The problem with doing this is that once you have a reasonable cache on which to compute optimal filesystems numbers, you then have to blow it all away and start again! So, here are the values I used on a 4.3 GB disk totally dedicated to the Squid proxy... -b 8192 -f 2048 -i 12288 The 'b' of 8 KB allows 1 inode to map up to 96 KB of disk which is more than adequate for HTTP objects and not too small for FTP objects while the # of inodes on the 4.3 GB disk with -i = 12 KB comes to about 350,000 which again is adequate for the 230,000 - 240,000 objects in the cache! If anything, you could make the frag smaller and possibly up the -i to 16 KB but be careful about running out of inodes... If the 2 x 3.1 GB disks are all that you have on the machine, then they need to contain the system files, swap space, the Squid log files, and the cache itself. I have 4 SCSI disks arranged as- 1 x 1 GB => system (/,/var,/usr) 2 x 1 GB => stripe set for Squid logs (-b 16384, -f 8192, -i 131072) 1 x 4 GB => Squid proxy cache There are also 3 swap partitions of 128 MB (384 total) on the 3 smaller drives as Squid can go 'strange' at times and swallow huge amounts of VM!!! We're still on Squid 1.1.3 as I havn't has the time to upgrade to 1.1.9 so some of the VM problems have probably been fixed allowing less swap space. The machine itself has 96 MB of RAM now and will have 128 MB shortly. Tony