On Tue, 20 Mar 2001 at 2:25pm, Steve Lord wrote
> p.s. Let us know if these numbers make things go faster!
>
OK. ;) The answer is a bit, but not really.
I verified with the vendor that, indeed, the stripe size specified on the
RAID system is blocks/disk. So, I upped the stripe size to 128
(512byte) blocks (I'll be dealing mostly with large files), made two
filesystems, and formatted them as following:
[jlb@philip jlb]$ sudo mkfs.xfs -f /dev/sdb1
meta-data=/dev/sdb1 isize=256 agcount=134, agsize=261319
blks
data = bsize=4096 blocks=35016700, imaxpct=25
= sunit=0 swidth=0 blks, unwritten=0
naming =version 2 bsize=4096
log =internal log bsize=4096 blocks=1200
realtime =none extsz=65536 blocks=0, rtextents=0
[jlb@philip jlb]$ sudo mkfs.xfs -f -d sunit=128,swidth=896 /dev/sdb2
meta-data=/dev/sdb2 isize=256 agcount=134, agsize=261328
blks
data = bsize=4096 blocks=35016704, imaxpct=25
= sunit=16 swidth=112 blks, unwritten=0
naming =version 2 bsize=4096
log =internal log bsize=4096 blocks=1200
realtime =none extsz=65536 blocks=0, rtextents=0
[jlb@philip /mnt]$ sudo mount -t xfs /dev/sdb1 /mnt/raid1
[jlb@philip /mnt]$ sudo mount -t xfs /dev/sdb2 /mnt/raid2
I benchmarked with bonnie++ and lmdd. If you want me to run any more
tests, just let me know. I'll still be playi^Wtesting for most of the
rest of this week. It's going to be a shame to let the grad students at
this thing...
For the record, the RAID is attached to a Dell Precision 410 with one PII
450 and 512MB of RAM. It's attached via an Initio INI-A100U2W card (Domex
branded). I'm still running 2.4.2-XFS pulled out of CVS on 3/13. For
those who don't recall the rest of this (week-old) thread, the RAID is
made up of 8 5400RPM 80GB Maxtor disks in RAID5 and appearing as a SCSI
disk to the host.
Here are the results (averaged over three runs):
[jlb@philip jlb]$ pwd
/mnt/raid1/jlb
[jlb@philip jlb]$ bonnie++ -r 512 -n 64
Version 1.01 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
philip.egr.duke. 1G 5058 98 30302 35 10498 16 4609 84 26341 22 133.8 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
64 1299 89 88532 98 2185 90 1239 84 99707 98 782 44
[jlb@philip jlb]$ pwd
/mnt/raid2/jlb
[jlb@philip jlb]$ bonnie++ -r 512 -n 64
Version 1.01 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
philip.egr.duke. 1G 4962 97 32775 38 9669 15 4498 84 26304 21 134.6 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
64 1144 79 88914 96 2196 92 1029 70 101379 95 585 32
[jlb@philip jlb]$ pwd
/mnt/raid1/jlb
[jlb@philip jlb]$ lmdd if=internal of=tmp bs=32k count=25000 fsync=1
819.2000 MB in 31.2588 secs, 26.2070 MB/sec
[jlb@philip jlb]$ lmdd if=tmp of=internal
819.2000 MB in 32.1173 secs, 25.5065 MB/sec
[jlb@philip jlb]$ pwd
/mnt/raid2/jlb
[jlb@philip jlb]$ lmdd if=internal of=tmp bs=32k count=25000 fsync=1
819.2000 MB in 31.3305 secs, 26.1470 MB/sec
[jlb@philip jlb]$ lmdd if=tmp of=internal
819.2000 MB in 32.3775 secs, 25.3015 MB/sec
Please note that I am *not* complaining about these numbers. I'll
take 26-30MB/sec any day. But if you have any more suggestions or want me
to do any other tests, let me know. Thanks!
--
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University
|