F@H on WLCG resources

Europe/Zurich
Julia Andreeva (CERN)
Description
In case of problems with Vidyo, Zoom link is: https://cern.zoom.us/j/6910698615
 

Attended:

Alessandra Forti, Costin Grigoras, David Michael South, Felice Pantaleo, Federico Stagni, James Catmore, Elizabeth Sexton Kennedy, Josep Flix, Maarten Litmaath, Julia Andreeva

Answers to the questions after ALICE presentation

Efficiency 85-90%, high success rate. Saving intermediate results has positive impact on the efficiency and success rate. Jobs get killed after 15 minutes if there is no new payload.

Alice did confirm with the sites that F@H payloads could be submitted to a given site. No objections received. Sites are willing to participate in this activity

Answers to the questions after ATLAS presentation

Resources used for the activity: Resources of some WLCG sites which agreed to participate and volunteer (personal) resources. Some fraction is GPUs.

Both ALICE and ATLAS are requesting to change the name of the monitoring page from CERN to something corresponding to the scope of the activity

CMS

F@H payloads were successfully run on HLT using Singularity, up to 3K. Exercised necessary machinery. Were not sure whether there would be enough payloads provided by F@H, therefore were thinking about possibility to run other applications.

LHCb

Is not planning to integrate workload management system with F@H. Want to understand better where this activity is going. Consider possibility to start with HLT farm with local submission. No integration with workload management system then would be required. LHCb experts are contributing to fight with COVID-19  providing technical expertise for DIRAC integration with various medical applications.

Overall, considering the scale of the activity, only small fraction of the WLCG resources (less than 5%) is foreseen to be involved. No effort is required from the sites. Experiment workload management systems are integrated with F@H activity and therefore submission and processing are transparent for the sites. All hard work has been done by the experiment experts. However, sites should be informed and agree to accept F@H payloads.

Actions

1). WLCG Operations Coordination to follow up on renaming monitoring page

2). WLCG Operations Coordination to contact  sites (despite the fact it has been already done by ALICE and ATLAS) and asking them whether they have concerns related to funding agencies and therefore can not accept F@H payloads

3). Prepare small session at the next GDB

 

There are minutes attached to this event. Show them.