[Openbenchmark] notes from today

[faibish_sorin at emc.com]

Per your email I will need some time to present the sfs netmist benchmark at the LUG. How do I get a spot for the presentation as you mention in the email. Who will be able to schedule this during the next LUG meeting? Thank you for your advice

/Sorin

-----Original Message-----
From: openbenchmark-bounces at lists.opensfs.org [mailto:openbenchmark-bounces at lists.opensfs.org] On Behalf Of Vanderbilt, Richard
Sent: Friday, March 23, 2012 5:09 PM
To: openbenchmark at lists.opensfs.org
Subject: [Openbenchmark] FW: notes from today

3/23/2012
OpenSFS Benchmarking Workgroup 

Participants:
Sarp Oral ORNL Richard Vanderbilt Netapp Henry Newman Instrumental Neil Martin Instrumental Cory Spitz Cray Frank Leers DDN, Branislav Radovanovic   Netapp, Eugene Birkine Xyratex, Stephen Simms Indiana Univ., William Arcand MIT, David Bester MIT, Galen Shipman ORNL, Rich Moore,  John Hammond TACC,  Richard Henwood  Whamcloud,  Sorin Faibish  EMC,  Liam Forbes Alaska EDU,  Dave Vasil DDN.  (I apologize if I missed anyone).

Notes: OpenSFS board has not yet approved the working groups charter.  We expect them to approve the charter this Tuesday at the next board meeting.
We had several new members join.  For the benefit of these new members we briefly discussed our charter and discussed where we are in the program.
One focus of this working group is to identify various I/O workloads in HPC environments primarily Lustre environments.  To standardize on the collection of data, this group Is assembling a questionnaire to send out to the Lustre community. The questionnaire is intended to retrieve information about their Lustre environment including: I/O workloads, configurations, Client/application information and the markets they serve.

Today, we passed around the first Draft of the questionnaire  (see attached document).  The initial draft of the questionnaire was generated by Henry Newman and Neil Martin from Instrumental and Sarp Oral from ORNL and Richard Vanderbilt from NetAPP.  We are asking the members to review the questionnaire and edit as they deem appropriate.  Our next meeting will review the edits and walk through the questionnaire. Those members in the Lustre user community, we asked if they would be kind enough to begin answering the survey.  These members can act as a pilot for sharpening the survey.

Members were asked on the call if they knew of any data collection tools used to collect the data we are seeking.  (Someone volunteered their tool but I'm sorry I didn't get your name.)  These tools should be freely available and preferably easy to deploy.  There were some tools mentioned from Berkeley Livermore Labs that could also be a starting point.

Sorin Faibish from EMC is proposing the use of a benchmark from SpecFS.  This benchmark tool can generate different workloads per profile.  Sorin is working on generating some profiles to emulate Lustre workloads.  Sorin, requested time at the face to face meeting at LUG to present this benchmarking tool.  We need to understand these tools better to make an informed decision.  Sorin, has provided more information in a subsequent email.

Neil and Henry will be working on a list of definitions and abbreviations used in our working document.

Once we become an official working group, our first requirement is to create a white paper defining in greater detail the objectives of this group, the content we plan to create, the schedules and progression outline.  I will begin working on this for next week's meeting.

I would like to thank Neil, Henry and Sarp for developing the initial draft of the questionnaire.  If you have any questions, feel free to send them to both Sarp and/or myself.

Kind Regards:

Benchmarking Work Group

Richard Vanderbilt - Netapp
1(925)872-4399 mobile
Co-Chair

Sarp Oral - ORNL
Co-Chair
1(865)574-2173