If you open https://github.com/VI4IO/io-500-dev/tree/master/proposal-draft/kaust-geor...
the readme file was updated.
So first of all, the file startup-auto-detect.sh will execute many instances of benchmarks
in order to find the appropriate parameters for you, as we want the benchmark to last for
a minimum duration. That’s why in the Readme file after the execution of the auto-tuning
tool you have the parameters that you need to define to the main benchmark, (see readme
In the startup-auto-detect.sh we have the paths
Which means you need to install the benchmarks IOR mdtest md-real-io and declare the
I assume you have already IOR.
Of course, change the core allocation as you prefer and define enough time limit, totally
it depends on the allocation and on the system, so I can not provide any magic number,
however, better more time than less. ☺
Don’t forget to read for the “parallel” find if you are interested to use it, you need to
use the script compile.sh from the main folder to compile it and to adjust the main script
(it is not used during auto-tuning), there are again instructions in the
If you have any issue contact us.
My final advice is as you do clone of the git account copy one of the site’s folder, give
your site’s name and work inside there (when you copy into your site’s name delete the
data of the results folder only). Then you could commit to github your site’s files when
you have finished.
George Markomanolis, PhD
KAUST Supercomputing Laboratory (KSL)
King Abdullah University of Science & Technology
Al Khawarizmi Bldg. (1) Room 0123
Kingdom of Saudi Arabia
Mob: +966 56 325 9012
Office: +966 12 808 0393<tel:%2B966%2012%20808%200683>
From: IO-500 <io-500-bounces(a)vi4io.org> on behalf of Julian Kunkel
Date: Wednesday, 23 August 2017 at 7:08 AM
To: Osamu Tatebe <tatebe(a)cs.tsukuba.ac.jp>
Cc: "io-500(a)vi4io.org" <io-500(a)vi4io.org>
Subject: Re: [IO-500] More testing
Please have a look at the individual starter scripts provided in the directories for Dkrz
or Kaust: *.slurm
The variables are defined there together with the parameters for the resource manager.
Basically there are two runscripts in each dir, one to run the io500 benchmark and one to
help determining the appropriate parameters to meet the 5 minute creation phases.
We will extend the descriptions once we have a few more runs.
If you have any question do not hesitate to ask them.
Am 23.08.2017 03:32 schrieb "Osamu Tatebe"
We are looking at the git repository. It seems io_500_core.sh
is a benchmark program, however, several variables should be
correctly defined. It would be helpful if there are some
documentations about the codes. These scripts are readable
but it helps to understand correctly.
Osamu Tatebe, Ph.D.
Center for Computational Sciences, University of Tsukuba
1-1-1 Tennodai, Tsukuba, Ibaraki 3058577 Japan
From: John Bent <firstname.lastname@example.org<mailto:email@example.com>>
Subject: [IO-500] More testing
Date: Tue, 8 Aug 2017 11:20:29 -0600
Now that George has run successfully and demonstrated that the benchmark suite
is working correctly at Kaust, it is time to ask our next batch of volunteers
At ISC, we had 6 volunteers to do this. I've added you six on the bcc line;
please join the mailing list if you haven't already (
). And please reply with
any questions you have about how to do the benchmarking.
IO-500 mailing list
This message and its contents including attachments are intended solely for the original
recipient. If you are not the intended recipient or have received this message in error,
please notify me immediately and delete this message from your computer system. Any
unauthorized use or distribution is prohibited. Please consider the environment before
printing this email.