lkml.org 
[lkml]   [1998]   [Jun]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: problem: linux and (really) fast I/O
Date
Davide Rossetti <rossetti@apemaia.roma1.infn.it> wrote:
> DISKS="/dev/md0 /dev/md1 /dev/md2 /dev/md3 /dev/sdb1 /dev/sdd1 /dev/sde1 /dev/sdf1"
> time dd if=/dev/zero of=$dsk bs=$blksize count=$count

The problem is that when using the raw (?) disk interface, you will
see at the SCSI analyzer that delay between commands is around
milliseconds. This probably because

write("/dev/sdX", size); -> copied in buffers
-> sent to SCSI layer as maybe one 64k block
-> command sent to drive. When command finished,
at that time the user process is wakened up
and THEN ONLY it fills again the buffer, etc.

At least it's my understanding. NB this has been seen on the SCSI
bus at 2.0.x times, so maybe it has improved.

Try your benchmarks on a *filesystem*, you will see it will become
better (I saw upto 28 MByte/s on a P200 64 MB with Bonnie ext2
-b 4096 Adaptec 2940UW with a SCSI-to-SCSI RAID0 array with
two IBM DGVS drives).

I have not tried md stripping of DGVS drives yet (I have tried
slower DCAS and it *does* double the speed, and especially the
seeks).

Even better someone has started a high performance file system suited
for video applications under Linux (can't remember the name, though).

Mail me if you want the Bonnie benchmark program and some more
tips.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu

\
 
 \ /
  Last update: 2005-03-22 13:43    [from the cache]
©2003-2011 Jasper Spaans