On 24.04 00:58, Charlie Johnson wrote: > >> The company I'm working for uses squid for caching and access control. > >> We use 3 Dell 2650s for squid, with a 3 disk (U320 10K RPM) RAID-0 > >> (stripe set) for the cache partition on each server. > >What OS? > >It's recommended to split that stripe into separate drives. Squid does > >not benefit from striping. Squid automatically splits the load on the > >available drives (cache_dir), so striping only makes long term > >maintenance of your system harder as any change to the cache would mean > >loosing the whole cache. > The OS is Linux 2.4.29 (Debian 3.0_R4) > What is the recommended size of each cache_dir (disk)? > Still, with several smaller independent disks for the cache_dirs, what > filesystem is recommended for this use? http://www.squid-cache.org/Doc/FAQ/FAQ-4.html#ss4.14 it is the same for every drive, and it doesn't matter if you use separate drives or strip from this point of view. > >> What file system is recommended for use on the cache partitions? > >> We have tried with ext2, xfs and reiserfs 3.6. > >> > >> First we thought using ext2 (no additional configuration) would be a > >> good idea, since there is no journaling etc. The performance sucked. > >What cache_dir type did you use? aufs is recommended for linux. > >The default "ufs" cache_dir type by design won't perform in higher > >loads as each I/O operations blocks the whole Squid process.. That's it. first set up aufs, then try playing with different filesystems. Some tests showed that ext2/3 behaves better than e.g. reisersfs for squid cache. I didn't try but I think ext3 should be good (journalling slows it down a bit, but in case of crash you won't have to wait long time until fsck fixes all). > cache_dir diskd /var/spool/squid 65536 64 1024 the count of first-level and second-level directories should be close to eash other and I don't recommend you increasing the second one (the default is '16 256'). The number of directories really used depends on number of object stored in the cache and therefore on maximum_object_size. I've used '256 256' on 30GB drive, but only 64 of first-level directories got used, so I decreased it to '64 256'. The average object size is ~13 KB so I had ~2 milions of objects on that dir. If you have 70 GB of space, I'd use '128 256'. But you are going to split that up to 3 disks, each of lower size, right? In such case, '32 256' should be enough. Our caches, -- Matus UHLAR - fantomas, uhlar@xxxxxxxxxxx ; http://www.fantomas.sk/ Warning: I wish NOT to receive e-mail advertising to this address. Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu. They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety. -- Benjamin Franklin, 1759
