90 likes | 182 Views
Summary of Parallel Session on Offline Data Processing J. Urheim Sunday, July 14 2002. Initial Goals of Processing. Work out the technical aspects FNAL batch farms, set up infrastructure (A. Sousa, H. Rubin) Offline software: performance, robustness
E N D
Summary of Parallel Session on Offline Data Processing J. Urheim Sunday, July 14 2002
Initial Goals of Processing • Work out the technical aspects • FNAL batch farms, set up infrastructure (A. Sousa, H. Rubin) • Offline software: performance, robustness • Provide output that is useful for collaborators • Try to process all events, minimal filtering • Set up to do nearly real-time processing for quick turnaround • Output should be useful for further development of reconstruction: should contain raw data + candidates • Output should be useful for higher-level data analysis output ntuple(s)?
The MINOS Farm Processing Infrastructure [A. Sousa] Work done in collaboration with Howie, Liz and Jon. Overview: • Develop an infrastructure for the MINOS Production Data Processing • Automate the Production Data Processing setting up for two modes of operation: • Run through the full data sample • Run on data files as they become available • System to be setup using the Fermilab Fixed Target Farm (fnsfo) [NuMi, KTeV, HyperCP(E871)...] http://mixcoatl.phy.tufts.edu/asousa/production/production.html
The MINOS Farm Processing Infrastructure [A. Sousa] The Farm Processing Block Diagram
Benchmarking of Recons’n s/w[R. Lee] • It is possible to keep up with data taking • Both R. Lee & Indiana are currently doing this. • Has led to some performance improvments • But will slow down when B field turns on… some benchmarking results from Roy
Calibration Components [J. Musser] • Offline Framework • Timing Calibration • T0 • Timewalk • Charge Calibration • Drift Point • Non-Linearity • Inter-Strip Calibration/Response Flattening • Inter-Detector (Stopping muons) • CalDet Energy Calibration • Alignment
CalDet Data Processing • CalDet requirements differ • Main goal is to put data in form so as to reduce overhead on additional offline event reconstruction. • triggering, applicat’n of calibrations, etc. • Benchmarks from Mike Kordosky:
Questions & Issues • What to run? • DigitListModule - yes! • DeMuxCosmicsModule - yes! • TrackSR et al. - yes! • BubbleSpeak - yes! • what else ? • Which candidates do we write out ? • Do we write out an ntuple ? • Do we do any filtering ? • non-snarl records ? • events with fewer than N planes with hits ? N = ? • Multiple output streams ? • Event classification: i.e., stopping muons, fully contained events, multiple muons, etc…
Effort needed ! • “Offline Monitoring” process(es) • for quality control of reconstruction, calibration etc… • Event statistics go into event header record? • what else ? • Loose event classification software • Good projects for graduate students/postdocs