Brian wrote: > 1000-2000 concurrent users (8000 machines total, not all of them will be hitting > the internet all the time, most don't need it, like servers, etc), 6 T1s. > that's not an insignificant amount of traffic anyway. :)
Assuming concurrent users translates into a large number of requests per second (which squid measures) and traffic (MB/sec), your needs are WAY bigger than our little shop - hopefully some larger enterprise level proxy admins can help you. There has been lots of sizing/server config information posted and I know a lot of guys use multiple proxies (cache_peer). If your concurrent users translates into more than 200req/sec (depending upon your box/disks/RAM) then you may need more than one squid box. Just curious so I checked google and here are some interesting comments on max sizing: http://groups.google.com/groups?threadm=933r8s%242ra0%241%40FreeBSD.csie.NCT U.edu.tw http://groups.google.com/groups?threadm=9noqlu%241mf3%241%40FreeBSD.csie.NCT U.edu.tw Both threads are worth reading in their entirity. P.S. Here is a representative comment on manually setting async IO threads: http://groups.google.com/groups?selm=admnqh%241f2o%241%40FreeBSD.csie.NCTU.e du.tw Others just said adding more per disk is ok, squid defaults to 16 per disk/cache_dir. adam
