> so, you see a major inprovement in disk seek overlapping, don't you ?
Definately! Then again I use 8k chunks...
Personalities : [1 linear] [2 raid0] [3 raid1] [4 raid5]
read_ahead 128 sectors
md0 : active raid0 sda1 sdb1 4192896 blocks 8k chunks
/dev/sda:
Timing buffered disk reads: 32 MB in 4.86 seconds = 6.58 MB/sec
/dev/sdb:
Timing buffered disk reads: 32 MB in 4.85 seconds = 6.60 MB/sec
/dev/md0:
Timing buffered disk reads: 32 MB in 2.95 seconds =10.85 MB/sec
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
Vendor: SEAGATE Model: ST32550W Rev: 8104
Type: Direct-Access ANSI SCSI revision: 02
Host: scsi0 Channel: 00 Id: 01 Lun: 00
Vendor: SEAGATE Model: ST32550W Rev: 8104
Type: Direct-Access ANSI SCSI revision: 02
> In real life those are not zeroes. they are data from/to our supercomputer
> memories. And in our QCD(Quantum Cromo Dynamics) simulations, there are
> HUGE amount of data to be read/written.
Sounds like fun! You might be better off with chunk size < buffer
size. (just a pseudorandom guess really) Then again is fs even threaded
in 2.0?(can any fs guru verify this) I'm on 2.1.106ac3 using a 2940w
which seems concrete even with SMP+RAID+realtek 8029<g>.
- Chris
--------------------------------------------------------------------------------
Chris Atenasio (chrisa@ultranet.com) -- Friends dont let friends use Windows.
Send me mail with subject "send pgp key" or "word of the day" for autoresponse.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu