Please, excuse the long, confused, message. I am thinking I should do something that will make the web seem faster on the machines using Squid as a proxy server. I am using Squid version 2.1 (The actual number of the Release doesn't display on my Linux box instead I just get "Squid Cache: Version 2.1.RELEASE" when I run "squid -v") Should I be running an httpd accellerator? Would things be better if I tried to be a Transparent Proxy? Should I try to run Squid on another machine and create a cache hierarchy? How could I tell? Should I go and get a faster computer to run Squid or do I just need to configure it differently? I am pretty ignorant of what the differences are. [I spent a lot of time learning how to configure a computer as a firewall with ipfwadm and ipchains, simply because I didn't fully understand the differences between ip-masquerading, ipforwarding, and proxy-seving. Useful stuff to learn, but I could have simply run squid out of the box without even rebuilding the kernel and achieved what I wanted to do.] I am trying to improve Internet accessiblity for a small public library by using a proxy server. Our building has a functional LAN with CAT 5 wire on each of the three floors and fiber between the floors. We are part of the Southern Adirondack Library System (SALS). We are connected to them through a partial T1 (8 - DSUs) and they function as our ISP so this is also how we are connected to the internet. In effect inside the building we are connected through an ethernet 10baseT LAN. Many of the services, including DNS and DHCP are being provided by SALS. My feelings are that if I provided DNS and let the proxy-server machine function as the Gateway and ran a cache accelerator, performance would improve but eventually the machine I'm using would be overtaxed and things would deteriorate. I am not sure of any of this. At present Squid is only serving 3 machines I put out in the reference section of the library. The disk partition with the cache does have at least 200MB of disk space to spare at the moment. I am running Squid on an HP Vectra with a 75MHz pentium with RedHat and Linux Kernel 2.2.1. It has 24M of RAM. (I upgraded other machines and had lots of left over 4M RAM chips to stick in the 6 slots.) I am running some other experiments including an apache web server on this machine. We set all the machines in the library to have the same home page on a server outside the library. So the accellerator would probably only speed up access to that one page, if I mirrored it inside. How would I tell if it would make a difference? I am sure I will be able to overtax this server eventually. I need to get a good idea of how much it can handle, so I can tell what I would need to get it to work well or what I would need to get to replace it. Except for trying different things with acls and authentication it is still pretty much a default installation. I altered the cache and log directories so that they are symlinks. lrwxrwxrwx 1 squid squid 16 Feb 24 15:12 cache -> /var/cache/squid lrwxrwxrwx 1 squid squid 14 Feb 24 15:12 logs -> /var/log/squid This use of the "/var" has to do with a previous job as a systems administrator where I realized that the "/var" directories are so named for a reason and that crashes are less servere if logs (though of course there should be a log host kept inaccessible behind a firewall) and files that change size dramatically are kept there. Note that I keep such things on separate partitions and also have space limitations. Filesystem 1024-blocks Used Available Capacity Mounted on /dev/hda1 118965 29119 83699 26% / /dev/hda9 148804 71480 69640 51% /home /dev/hda5 644431 227269 383875 37% /usr /dev/hda6 476194 132558 319043 29% /usr/src /dev/hda7 446421 81649 341716 19% /var /dev/hda8 446421 9692 413673 2% /var/log -- Josh Kuperman Saratoga Springs Public Library [EMAIL PROTECTED] 49 Henry St 518.584.7860x211 Saratoga Springs, NY 12866 http://www.library.saratoga.ny.us
