On Mon, 20 Mar 2006, RS RS wrote:> Hi, > I''m running the sgpdd-survey program. That seems to want SCSI > drives, but I have IDE drives (ATA). I was able to modify the script to work > with > a partition on my IDE drives. Has anyone else done this?sgp_dd is designed to use sg devices, so if sg_map doesn''t give you a mapping you can''t use it, period. The whole point of sgpdd-survey is to do IO in a similar manner of what''s lustre is doing, and that means bypassing most of the standard IO layer. /Nikke -- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Niklas Edmundsson, Admin @ {acc,hpc2n}.umu.se | nikke@acc.umu.se --------------------------------------------------------------------------- Black holes resulted when MS tried to beat a deadline =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Hi,
I''m running the sgpdd-survey program. That seems to want SCSI
drives, but I have IDE drives (ATA). I was able to modify the script to
work with
a partition on my IDE drives. Has anyone else done this?
I am getting some errors, however, and the performance seems very low.
I''ve seen performance as high as 43 MB/s, but that not consistently.
The
disks are Maxtor 6Y080L0.
Can anyone give me advice about getting rid of the errors and improving the
performance? The output looks as follows:
total_size 1048576K rsz 1024 crg 256 thr 256 write
./sgpdd-survey: line 123: 497 Terminated sgp_dd $inf $outf
${skip}=$((1024+j*blocks)) thr=$((thr/crg)) count=$count bs=$bs bpt=$bpt
time=1
2>${tmpf}_${i}_${j}
./sgpdd-survey: line 123: 560 Terminated sgp_dd $inf
$outf ${skip}=$((1024+j*blocks)) thr=$((thr/crg)) count=$count bs=$bs
bpt=$bpt time=1 2>${tmpf}_${i}_${j}
2 failed read 17.51 MB/s 256 x 0.07 = 17.09 MB/s
total_size 1048576K rsz 1024 crg 256 thr 512 write ./sgpdd-survey:
line 123: 4118 Terminated sgp_dd $inf $outf
${skip}=$((1024+j*blocks)) thr =$((thr/crg)) count=$count bs=$bs bpt=$bpt
time=1 2>${tmpf}_${i}_${j}
./sgpdd-survey: line 123: 4269 Terminated sgp_dd $inf $outf
${skip}=$((1024+j*blocks)) thr=$((thr/crg)) count=$count bs=$bs
bpt=$bpt time=1 2>${tmpf}_${i}_${j}
2 failed read 17.38 MB/s 256 x 0.07 = 17.09 MB/s
Can anyone say why? I am getting the same error on multiple
machines, so I don''t think that it is the disk itself. Also, I
don''t get
any
errors on the first few iterations of the program.
Thanks.
-Roger
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it''s
FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/