I would suggest you analyze the cache.log to see which objects are responsible for miss, size of object, TTL of the object, do you have many objects with query string. If so, should they be cached? Do you want to cache object which does not have TTL? (currently your squid.conf does not cache this type of object). It is better to start with understanding your traffic, user behavior and what you can compromise in term of object's freshness to gain saving. The higher hit rate and longer the object stays in cache, the better of saving. Regards, Khanh -----Original Message----- From: Errol Neal [mailto:eneal@xxxxxxxxxxxx] Sent: Wednesday, October 18, 2006 12:28 PM To: squid-users@xxxxxxxxxxxxxxx Subject: Fine Tuning - Suggestions If you would.. As an experiment, I implemented SQUID 2.6 on my network to see if we could reduce our bandwidth utilization. I ran it for about a week in production and after getting through my other work, I was finally able to analyze my access logs using squeezer2 and see how we fared. According to the report that reviewed, my Hit Cache was about 31% and we managed to save about 9% on our bandwidth utilization. Frankly, I was hoping for squid to make a bigger impact than that, but I am aware that this is probably going to take some more tuning. My question is how do I go about doing that? I'd like to increase see a much bigger hit cache ratio and a larger bandwidth savings. Also, should I consider running squid for a longer period of time before performing any kind of analysis? I'm posting my squid.conf file. Any advice, chastisement, rebuke, etc would be appreciated. TIA, Errol Neal ### acl QUERY urlpath_regex cgi-bin \? no_cache deny QUERY cache_mem 128 MB refresh_pattern . 0 20% 4320 refresh_pattern -i jpg$ 0 50% 8640 refresh_pattern -i png$ 0 50% 8640 refresh_pattern -i gif$ 0 50% 8640 refresh_pattern -i swf$ 0 50% 8640 refresh_pattern -i js$ 0 50% 8640 refresh_pattern -i css$ 0 50% 8640 refresh_pattern -i pdf$ 0 50% 8640 refresh_pattern -i flv$ 0 50% 8640 refresh_pattern -i avi$ 0 50% 8640 cache_replacement_policy heap LFUDA memory_replacement_policy heap LFUDA maximum_object_size 150096 KB minimum_object_size 0 KB fqdncache_size 4096 ipcache_size 4096 logfile_rotate 0 access_log /usr/local/squid-2.6/var/logs/access.log cache_log /usr/local/squid-2.6/var/logs/cache.log cache_store_log /usr/local/squid-2.6/var/logs/cache_store.log half_closed_clients off cache_swap_high 95 cache_swap_low 90 cache_dir aufs /var/squid/cache1 228828 25 256 cache_dir aufs /var/squid/cache2 228828 25 256 buffered_logs on http_port 3128 transparent acl localnet src 172.16.100.0/255.255.252.0 acl localhost src 127.0.0.1/255.255.255.255 acl CONNECT method CONNECT acl all src 0.0.0.0/0.0.0.0 http_access allow localnet http_access allow localhost http_access deny all ### __________________________________________ Errol Uriel Neal Jr. Sr. Network Administrator DFI International, Inc. 1717 Pennsylvania Ave NW, Suite 1300 Washington, DC 20006 Tel (202)452-6955 Fax (202)452-6910 eneal@xxxxxxxxxxxx www.dfi-intl.com
