WLCG throughput call

Wednesday 27 Jul 2016, 15:00 → 16:00 Europe/Zurich

31/S-023 (CERN)

0) Agenda Review/Update, News 1) perfSONAR status - Issues to report for 3.5.1 - WLCG deployment/operations status 2) WLCG network throughput SU tickets 3) Focus Topic: Re-organisation of the WLCG-wide meshes 4) Round-table about throughput, network, monitoring, data transfer and new issues to track 5) AOB and next meeting

=============================================

NOTES: North American Throughput Meeting --- July 27, 2016

=============================================

 

Attending: Dave Lesny, Mark Feit, Shawn McKee, Garhan Attebury, Andy Lake, Saul Youssef, Horst Severini, Ilija Vukotic

Excused: Jason Zurawski

 

 

0) Agenda review -- No updates

 

1) perfSONAR status

     i) New network problems  -  No new issues reported. OU/LU net issues looking much better after Joel Snow reported the problem to OneNet and OneNet   opened tickets with ESnet/Internet2. Unknown what the solution may have been and Horst will let us know if details are provided.  

Saul reported timeouts to NET2 have disappeared. Here's a link to a history of the problem.

          http://egg.bu.edu/NET2%7binf:NET2%7d/gadget:Studies/section:report/2016-06/McGill_timeouts_to_NET2/  

 

    ii) Updates on perfSONAR, MaDDash, mesh-config, etc.  Andy reported that we are on track to have a release candidate of perfSONAR v4 in September.

 

          Garhan asked about difference between maddash.aglt2.org vs perfsonar-itb.grid.iu.edu vs psmad.grid.iu.edu.   Shawn: maddash is prototyping instance (development/testing), perfsonar-itb is the OSG testing instance (for software in beta heading toward production) and psmad is the     "production" instance.   Garhan noted the prototype instances is missing hyperlinks on some of the USCMS sites.  Shawn noted the prototype instance has the toolkit URL on the rows and the new "report" URL on the columns.  Andy:  this is not the expected behaviour and he will look into both things.

 

2) Presentation on pScheduler  (Mark Feit) --- Slides attached to the emailed meeting announcement.

         Mark gave very nice overview of pScheduler.   

       

3) Round-table (new issues, items to share or discuss ?) Garhan reported the perfSONAR development node was "down" because of operator error. Shawn: plan 

       for the testbed is to test the next release when ready. PuNDIT also trying to use pScheduler 

 

       Ilija mentioned the network analytics work that is ongoing and we will have a presentation on it at the next meeting (see below).

 

4) AOB and next meeting.   Next meeting will feature Xinran Wang presenting on his and Ilija's network analytics work.   Meeting date TBD pending discussions with Marian about the European version of this meeting.

 

Send along any additions or corrections to these notes via email. 

 

Shawn