2012 Software Review Closeout Notes
From GlueXWiki
The final report is not yet available, but a draft was read at the closeout.
Generally, a very positive reaction for Hall-D. Some comments from Dave's notes taken during closeout:
- Future reviews might include talks from end users to relate their experiences
- We should be working towards large scale mock-data challenges sooner rather than later
- Standard profiling/evaluation tools should be integrated so that they are run routinely
- with new releases
- More attention on file discovery(??)
- Workflow tools are in conceptual stage
- Invest in transition to GEANT4 only when extra manpower is available or a physics motivation
- arises (this confirms of our current position)
- Accomodate natural peaks in CPU needs during analysis when forming requirements
Recommendations
- Plan a series of scaled tests using the LQCD farm
- Consider using ROOT for the DST file format
Moving Forward
Curtis sent around a similar summary, but added that we should probably organize a "Calibration Workshop" to start focusing more on specifics there.
Matt elaborated a bit more on what might be done for the data challenge:
I think the most important next step is to do a one day data challenge (or even repeat the 3.5 hour one) in the style that we will do the final analysis. Two key steps: * separate the reconstruction and analysis step -- we need to have JANA ready to read in the reconstructed DST format (that we haven't quite specified yet) and make analysis level C++ objects ready for the user analysis * In doing this we should experiment with how we manage files for batch submission. Also we can play around with different mechanisms to manage a user skim (EventStore or some equivalent that ROOT might have). We should have a goal of one large MC set that can be consumed by multiple users (e.g. 3 pi and 5 pi). This is the approach we will use with the final data.