Hi *Squiders*,
you wrote on 20 Feb 97 at 22:48 in newsgroup /mailinglisten/squid-digest
in "squid-users-digest Digest V97 #43":
> I'd suggest to use 1 physical disk (or stripe or raid) per cache dir.
> Calculate expected max number of objects in your cache, using average
> object size of about 15-20 KB, ie. TOTALSIZE/average="total objects".
> Then divide this by 256 which is optimal number of files per level 2
> dirs and you'll get total number of needed directories in both level
> 1 & 2. Numbers for each of these should be power of 2 (4, 8, 16, etc)
> Then, I'd suggest to select the lowest needed number for level 1, keeping
> level 2 smaller than 256 (make it 128). Then divide no of level 1 dirs
> by number of disks you have and you'll finally get needed numbers to
> put into config. If you have different sized disks, use smallest sized
> disk to calculate max size.
> All this is also noted in Release notes.
>
> All this is good for is to help OS make directory accesses faster,
> reducing no of directories helps to reduce needed cache size OS allocates
> for reducing physical disk io to find file in a directory.
Thanks a lot Andres, I think I understand it now !
But what I still like to know is:
What store_objects_per_bucket ?
(the comment in the config doesn't help me to understand - what is the
"store hash table" and what will happen when I change this value?)
If I make a new "mke2fs", what inode size do you recommend ?
Andreas
-- Andreas Reeh = A.Reeh@WaD.org * 51545 Waldbroel * Germany * PGP welcome More info, public key, geek-code and other email and talk addresses at http://www.wad.org/~andreas/ or http://home.pages.de/~andreas.reeh/Received on Fri Feb 21 1997 - 13:58:22 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:31 MST