MDC3: EbyE Physics Working Group

  1. BNL attendance - We expect about 15 persons (including BNL residents) from EbyE to be present at BNL and active during MDC3. Present office space is probably ok. I have solicited input from anyone not having peronal computing access (and therefore needing to use an xterm facility at BNL).
  2. Goals - We are emphasizing three baseline analyses: flow, global variables and two-point correlations. Flow plans to analyze 20k events (with and without HBT). The other two analyses will analyze 30k events with three types of correlation structure, and should also analyze the flow events as an exercise. In some form or another we plan to analyzed 50k events, perhaps more than once. As an intermediate step a miniDST of primary tracks should be formed, which is then the subject of further analysis. The result of this further analysis will be histograms and number tables (covariance matrices). We should be able to bring these analyses to the point of `physics' interpretation and presentation of results at 3 pm meetings - a full dress rehearsal for real data. The emphasis in this exercise is two-fold: exercise CAS infra in a major way and develop large-scale physics analysis techniques through to final interpretations
  3. Data volumes - for 50k evts about 1 TB of DST and 100 GB of miniDST. All subsequent volumes should be trivial in comparison.
  4. Chain - it would be interesting if the primtrk miniDST can be written as a common `chain' activity. The primtrk consumers as we have discussed are Spectra, HBT and EbyE. If several primtrk cut systems are in force it should be possible to mark tracks as passed by one or more of the three cut systems. The miniDST would contain all of the passed primtrks as a Boolean or. Different analyses would then use tracks with the appropriate cut flag. We then consider the tactical problem in which multiple analyses are touching a common large disk-resident data volume.
  5. Joint analysis between PWGs is possible and desirable, but perhaps not plannable at our present stage of preparation. We already have a joint analysis within EbyE, with three related but distinct major initiatives.
  6. Dhammika was our LSF expert. Jeff has used it in the past. In the next several weeks we will use the 18k Venus events to prototype the three analysis systems, learn how to use LSF and find out what other infra is missing. Jeff, Janet and Ron have experience running on CAS machines.
  7. One of the things to come out of this exercise should be knowledge about bottlenecks. We heard at CHEP that bottlenecks in BaBar analysis have been significant, and some require additional expenditure to solve. One wants to try to reduce bottlenecks and other performance-limiting aspects by good design and coordination as early as possible.

Text by Tom Trainor, Feb 15, 2000