1 of 3

BigPanDAmon@COMPASS experiment

Tanya Korchuganova on behalf of BigPanDAmon team

​

2 of 3

COMPASS experiment & PanDA WMS

  • COMPASS experiment is active since 2002
  • ~250 members, 24 institutes, 13 countries
  • Collects ~1.5 PB / year = ~20 PB in total
  • Data stored at CERN
  • Data processing and MC simulation �at CERN and JINR (Russia), �allocations at several HPCs: BlueWaters, FRONTERA, Stampede
  • Analysis - CERN lxplus or locally.

BigPanDAmon was adapted in 2017. �It took ~1 month and since November of 2017 it is in production.

Providing support and developing of new features on demand.

2

Adapted from ATLAS WMS:

Developed at COMPASS:

- ProdSys2

-JEDI

+PanDA

+Pilot

+Harvester

-Rucio (DDM)

+AGIS/CRIC

​

+BigPanDAmon

Production system

3 of 3

COMPASS BigPanDAmon setup

MySQL DB: a few actions needed despite using Django ORM.�Mostly related to specific features of Oracle that are not supported by ORM and used to aggregate metadata in scales of ATLAS production and analysis.

PanDA queue (computing site) level: no changes needed;

PanDA job level: a few minor changes done, e.g. configuration of the list of parameters to show;

Task level (set of jobs):

  • A separate DB instance and different table structures needed big changes to both backend and frontend sides.
  • No “dataset” abstraction caused massive plug-out of deeply embedded into the core of BigPanDAmon features

Campaign level (set of tasks): the custom “Production” module was developed.

3