I experience a strange phenomenon. If I boot linux 2.0.30 via disk I get the
following bonnie output:
-------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
1024 6873 96.3 14970 79.9 4744 39.2 9226 88.0 17184 55.8 57.7 2.2
If I boot the same kernel via lilo from harddisc I get:
-------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
1024 5210 82.7 9329 60.0 4683 49.5 8543 88.0 15956 55.5 74.4 3.1
Both tests were done several times, with results comparable to the ones
I show here.
My setuop is:
-Tyan Dual PPro motherboard (natoma chipset), with 2 CPUs (200MHz)
-256MB RAM
-Adaptec 2940 UW with 2xQuantum atlas 4GB UW
-linux kernel 2.0.30 with May8 adaptec driver
-The tested partition was striped on both harddiscs...
mdtab says: /dev/md0 raid0,8k,0,f6e2380e /dev/sda5 /dev/sdb5
The only reason I can imagine at the moment is that the adaptec bios does
something with the controller or the harddiscs before loading lilo
and the kernel when booting from harddisc.
Any other ideas?
I'm not sure which performance is better, as the reads are about the same and
when the writes are low the seeks are high and vice versa. But I think we
should get one consistent behaviour.
Matthias