-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Henrik Nordstrom wrote:
| Right, disk I/O is a major performance bottleneck for Squid. Not very
| much software tuning in Squid other than to enable aufs and the tuning
| knobs is mostly found outside of Squid
|
| 1. Number of drives used for cache
|
| 2. Type of filesystem used.
|
| 3. noatime and similar mount flags
|
| 4. make sure the L1 parameter in cache_dir is sufficiently large for
| your cache size.
I use this formula, found some where
cache_dir aufs /var/spool/squid C L1 L2
L1 = ((C*1000)/256/256/13)*2
L2 = 256
so with
C = 10000 => L1 = 22
or
C = 50000 => L1 = 110
is this formula reasonable ?
Before to use this formula I was using:
C = 50000 and L1 = 16
and believe me the performances were realy
bad.
Another question, if I want rise C and then L1
do I have to reinitialize ( delete and recreate )
the swap ?
Thank you
Regards
Gaetano Mendola
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAagTA7UpzwH2SGd4RAjEuAJ9zulcFZvL4Tcm24VsB384FiNbfvwCg3lG8
FRLl+Iah3IBQTxEowgopXns=
=6Mwq
-----END PGP SIGNATURE-----
Received on Tue Mar 30 2004 - 16:37:46 MST
This archive was generated by hypermail pre-2.1.9 : Thu Apr 01 2004 - 12:00:03 MST