On 8/05/2015 1:54 a.m., Veiko Kukk wrote: > > On 07/05/15 15:16, Amos Jeffries wrote: >> On 7/05/2015 11:16 p.m., Veiko Kukk wrote: >>> Is it possible to sort into cache_dir's based on filename (some regex >>> will do)? The result I'm trying to achieve is that certain files which >>> names are known, but size varies (and overlaps with other files) are >>> replaced with heap GDSF while all the other files are replaced with heal >>> LFUDA policy. >>> >>> Can this be done with squid? >> >> No. Replacement policy is a type of hash table for a cache_dir such that >> the single entry that is most obsolete can efficiently be found and >> discarded at the time point where cache runs out of space. The whole >> data structure for the cache_dir index is different for each mechanism. >> >> Why are you wanting this? What are you actually trying to do? >> >> Amos > > Hi, > > I'm trying to have different cache replacement policy for different > files, based on filename. To avoid bigger files pushing some smaller > files (more important files) out from cache in case of LFUDA and at the > same time have high byte hit rate for bigger files. I could achieve this > if squid could be configured to sort files based on filename/url into > different cache folders. There you go. Your problem statement is phrased in terms of sizes - "big files", "little files". cache_dir sectioned with min-size / max-size options using different replacement policy in each will do as close to what you need as is possible. PS. One can;t use "filename" because there is no such thing as "file" in HTTP. Only messages and objects with URLs. Any given URL can present a wide range of different sized objects. Any semblance to "files" is an illusion. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
