ADSM-L

Re: [ADSM-L] ITDT write speed tests vs. dd write speed tests

2010-04-09 22:07:13
Subject: Re: [ADSM-L] ITDT write speed tests vs. dd write speed tests
From: Wira Chinwong <wira AT SCSI.CO DOT TH>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Sat, 10 Apr 2010 09:06:00 +0700
I have ever tested about dd command. Speed up is limited by dd program. Just
try to test run one dd command by sending output to NULL. This will help you
to prove that the bottleneck come from Disk or Tape. If the transfer rate
result same as using "dd" to tape. Try to test run more than one "dd"
command to NULL in parallel. The result will tell you that there is
bottleneck on Disk or not. If transfer rate can increase, means that you are
facing the limitation of dd command.


Best Regards,


Wira Chinwong


-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Michael Green
Sent: Saturday, April 10, 2010 5:47 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: [ADSM-L] ITDT write speed tests vs. dd write speed tests

I've been playing with ITDT v4 for quite some time now. I installed it
on one of my production servers which is a hefty
IBM x3850, SLES10 SP2 x64, lin_tape 1.34. The drives are LTO4.
The FC fabric is 4Gb all along from the server to the library.

I'm getting mixed results that I cannot explain.
Using the "Full write" test from ITDT using Incompressible data and
256K transfer rate I consistently score above 100MB/sec write rates.
At the same time using simple dd method (with compression turned off,
same drive, same cartridge) as described in Richards ADSM Quick facts
(search for THROUGHPUT MEASUREMENT topic) I score no more than
45MB/sec.
The dd method is simple:
1. turn drive compression off (using ITDT or tapeutil)
2. run
# time dd if=/dev/zero of=/dev/IBMtape7 bs=256k count=40960.
(40960 by 256K is 10240MB). Now measure the time and divide 10240MB by
it. This test consistently yields ~45MB/sec in my experiments.

TAR'ing similar amount of data in a few big files yields about
40MB/sec. The underlying disks (a mirror of 2 146GB SAS 3Gb drives)
are well capable of providing 80MB/sec sustained.

Anyone tried doing similar tests? Anyone bothered playing with ITDT
read/write tests at all?
--
Warm regards,
Michael Green

<Prev in Thread] Current Thread [Next in Thread>