PandaReviewMar06

Panda Review March 8-9, 2006

US ATLAS internal review of Panda and Panda-based distributed analysis.

Charge:

Evaluate the progress of the PanDA project. Does the software design meet the ATLAS requirements? Are the physics requirements for the upcoming CSC being addressed in a prioritized manner? Evaluate the PanDA operation model and recommend strategies for its success. Compare PanDA with the overall ATLAS prodsys efforts. Review the interaction/integration of the ATLAS DDM software and grid middleware components in PanDA. Evaluate the integration of Dial into PanDA and recommend whether Dial should be maintained as an independent project. Evaluate the people resources (both program & base funded) and suggest optimization to ensure that a successful usable software tool that supports both production and distributed analysis needs is deployed for CSC.

Committee: D. Malon (chair), P. Calafiura, K. Cranmer, D. Liko, S. Mckee

Agenda with linked talks

Recommended resources to look at

• Panda Wiki

• Panda monitor, production overview, data browser

• pathena guide

• DIAL demo

• Data access tools, instructions

• Panda data management

• Panda code: CVS, LXR

• Panda database schema (Version 4 is current)

• Panda savannah (Not in active use yet)

• CHEP talks (attached) on Panda, US ATLAS production experience in Rome/DC2 production, and DQ2

• SciDAC-2 proposal (attached) for evolving Panda into an experiment-neutral workload management system for deployment on the OSG, through collaboration with Condor and CMS. Program to be undertaken if and only if the proposal is supported (ie existing scarce Panda manpower will not be diverted into this program, the deliverables of which are not on the US ATLAS critical path). The proposal provides some narrative on Panda, how Panda might make greater use of Condor (the areas of possible Panda/Condor integration discussed in the proposal are based on discussions with the Condor team in a visit by Panda people to Madison last October), and CMS interest in Panda-style "just-in-time workload management".

Review report

The review report is attached below.

Responsible: TorreWenaus