r2 - 27 Aug 2010 - 18:18:53 - RichardMountYou are here: TWiki >  AtlasSoftware Web > Minutes30Jul2010

Minutes30Jul2010 RAC Minutes, July 30, 2010

Members (*=present, #=apologies)

*Richard Mount, Kevin Black begin_of_the_skype_highlighting     end_of_the_skype_highlighting begin_of_the_skype_highlighting     end_of_the_skype_highlighting (Physics Forum Chair), *Jim Cochran (Analysis Support Manager), Alexei Klimentov (ATLAS ADC), Ian Hinchliffe (Physics Advisor), *Rik Yoshida (Tier3 Coordinator), *Michael Ernst (U.S. Facilities Manager), Rob Gardner (Integration Coordinator), *Kaushik De (U.S. Operations Manager), #Armen Vartapetian (U.S. Operations Deputy)

Ex-Officio: Torre Wenaus, *Stephane Willocq, Mike Tuts, Howard Gordon

Correction/approval of minutes of previous RAC meeting and core team meetings.

All were approved.

Summary of Operational Issues in the Last Month (Kaushik)


Panda Dynamic Data Placement has been in use since June 15 and has been very successful. The ADC distribution of ESDs and DESDs to US T2s has been stopped. Up to now, no reduction of AOD copies has been implemented. Rebrokerage (allowing jobs in the BNL queues to be moved to T2s as soon as they get the relevant data) will soon be in operation.

Even without rebrokerage, there appears to be no negative impact on user analysis. The BNL queues are very busy, but the distribution of work across the T2s is better than it was before PD2P was implemented.

PD2P has also been implemented with US help in France, and most recently in Italy. In France there are issues to be addressed related to France only having two small T2s.

Tadashi is making a small but important change to PD2P: when replication is triggered by activity in one cloud, the first replica will be distributed throughout ATLAS and will not be restricted to the cloud that triggered the replication. This should avoid having too many ATLAS jobs running in the US.

Stephane asked how popular were DESDs and AODs. Kaushik could not give exact numbers, but he noted that the usage of DESDs and AODs is rising, whereas ESD usage is (in a relative sense) declining. Off-grid Ntuple analyses are a major fraction of analysis activity. Thirty second jobs reading tens of kilobytes have been seen on the Grid, but are much better handled off the Grid.

Storage "Crises"

The filling up of BNL Scratch space was noted in the previous core-team minutes. After adding 10 TB and now that central deletion is running there is no immediate problem. It remains questionable whether the transient data storage used when BNL acts as a staging site for T2-T2 transfers is best handled by central deletion.

The SLAC storage crisis was also noted last week. The two contributing factors were 1) the poor handling of dynamically changing SpaceToken allocations by central deletion, which is becoming a problem for all T2s, and 2) the assignment of too much "archival" data to SLAC, which seems to have been the result of a misunderstanding somewhere in ADC. ADC have agreed in principle to implement a way (web service) to tell the central system about the status of dynamically managed space.

BNL Queues

Michael noted that BNL had implemented an automatic load-triggered reassignment of CPUs between production and analysis tasks. To ensure a rapid ramp up of production when necessary, there was a now floor of 500 slots assigned to production. Analysis jobs typically ran for 10-20 minutes, so no significant slow down of production ramp up is expected.

Summary of US T1+T2 pledged and unpledged resources (Michael)

This table had been circulated by Michael to US ATLAS computing management some time ago. The corresponding official ATLAS document of resource needs for 2010 - 2012 is https://twiki.cern.ch/twiki/pub/Atlas/ComputingModel/ATLAS_resource-needs-v16.pdf

Crystal Ball Session: how will/should US T1+T2 resources be used during the next 12 months? (All)

This discussion started from the resource information above. The US already had CPU resources exceeding the WLCG pledges, and was on track to have disk space that exceeded the pledge. The discussion centered on how to make sure that the beyond-pledge resources, that were in principle under US control, could deliver visible and vital value to US ATLAS physics analysis while avoiding any damage to the overall collaborative spirit of ATLAS. It was agreed that wider informal discussions were needed before writing a plan for the use of these resources.


Kaushik encouraged RAC members to participate in the ATLAS Physics Workshop of the Americas at Arlington July 9-11.

Action Items

  1. 5/7/2010: Richard, Create a web page summarizing the dataset distribution policies for the US resources.

About This Site

Please note that this site is a content mirror of the BNL US ATLAS TWiki. To edit the content of this page, click the Edit this page button at the top of the page and log in with your US ATLAS computing account name and password.


pdf 2011-pledges.pdf (17.7K) | RichardMount, 02 Aug 2010 - 13:54 | Installed US ATLAS Facilities (July 31, 2010) and Pledges for 2010-2012
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback