Here's the performance of bonnie on a 256M file,
using a 2P 64M x86 machine (same config as an earlier
test), using 2.4.0-test1 + XFS + bug fix to shrink_mmap()
that doesn't count page belonging to balanced zone as a try:
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
XFS 256 3450 95.8 6574 14.2 3148 8.4 2951 89.2 18623 29.5 176.5 4.1
XFS 256 3353 93.4 6509 13.1 2864 7.5 2966 89.5 18859 32.1 182.1 3.7
XFS 256 3468 96.3 6692 12.5 2753 7.1 2950 88.9 18447 29.6 181.6 4.8
EXT2 256 3434 94.8 6858 12.0 2949 6.1 3110 93.0 18713 21.3 174.9 3.3
EXT2 256 3421 95.0 6933 11.2 2628 6.0 3052 91.2 18569 22.0 176.0 2.5
EXT2 256 3474 96.5 6900 11.5 2824 6.1 3023 90.6 18103 24.4 173.5 2.9
While relative performance of XFS & Ext2 are same, absolute numbers
are still worse for block/char output than on 2.3.99pre2, both for XFS and
EXT2.
Here are the results for 2.3.99pre2 (sent out last week):
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
XFS 256 3633 100.0 7599 12.7 2906 7.4 2782 83.8 18480 27.4 181.7 4.3
XFS 256 3624 99.6 7614 13.8 2968 7.6 2776 83.6 18563 25.7 185.3 3.6
XFS 256 3625 99.7 7539 13.0 3014 7.9 2818 84.8 18557 29.4 183.4 4.9
EXT2 256 3626 99.9 7615 12.0 2932 6.1 3005 89.8 18495 22.7 176.7 3.3
EXT2 256 3632 100.0 7670 12.6 2920 5.9 2998 90.1 18483 22.4 179.6 3.3
EXT2 256 3631 100.0 7679 12.4 2931 6.3 3011 89.7 18478 24.0 187.0 4.2
There is more swapping in 2.4.0-test1 although not very much of it.
--------------------------------------------------------------------------
Rajagopal Ananthanarayanan ("ananth")
Member Technical Staff, SGI.
--------------------------------------------------------------------------
|