Presentation on theme: "Ma lisa - STAR collab mtg feb20021 QM02 production and mDST Straw-man QM02 production plan (to set scale) time limitations physics motivations storage."— Presentation transcript:
ma lisa - STAR collab mtg feb20021 QM02 production and mDST Straw-man QM02 production plan (to set scale) time limitations physics motivations storage requirements storage realities (summary) Common mDST looser & fatter than previously anticipated Iterate on straw-man QM02 production plan related issues: ordering of processing
ma lisa - STAR collab mtg feb20022 Straw-man QM02 production plan (Assuming no new calibrations show-stoppers) plan to migrate NEW PRO and begin QM02 production ~18 Feb ( just 19 weeks for production until end of 1 July) hope/desire homogeneous production (consistent reco software) for full production Use Jerome’s numbers for production rate (real time)… Au200Au central0.38 Mevts/wk Au200Au minbias1.4 Mevts/wk (*) pp minbias4.2 Mevts/wk (**) … and stats requirements for desired physics to estimate overall time machine duty factor 60% (*) x 0.7 if MinBiasVertex instead of ProductionMinBias (**) assumes 60% usable events
ma lisa - STAR collab mtg feb20023 Straw-man QM02 production plan - desired stats Au200Au central fullfield3.0 Mevts7.9 wk yield & crude spectrum high p T spectrum to ~10 GeV -bar K 0 -K 0 HBT Au200Au minbias fullfield3.0 Mevts2.2 wk v 2 of K 0, high p T v 2 Au200Au central halffield0.25 Mevts0.65 wk HBT: R (p T ), K- correlations low p T spectra Au200Au minbias halffield0.55 Mevts0.55 wk HBT: R (p T, ) pp triggers (incl hightower)?10.0 Mevts?2.3 wk spin analyses high pT reference K- correlations 13.6 wk ~ mid-late May workable if no major hiccups
ma lisa - STAR collab mtg feb20024 Storage requirements of QM02 production (ONE pass/version) can we continue with “business as usual?” I STAR DSTs (event.root files / no hits): AuAu central:( )Mevt x 1.8 MB/evt= 5.9 TB AuAu minbias:( )Mevt x 0.57 MB/evt= 2.0 TB pp10 Mevt x 0.1 MB/evt= 1.0 TB ~ 9 TB required can fit into the ~11 TB we should consider for QM02 prod So, yes, so far we could continue as usual, although capability for keeping multiple (e.g. pre- and post-QM02) or test productions available on disk would be severely restricted But what about the files that most PWGs usually actually use for analysis… ?
ma lisa - STAR collab mtg feb20025 HBT DST includes all of these (modulo details) Storage requirements of QM02 production (ONE pass/version) can we continue with “business as usual?” II The “big four” PWGs (EbyE, Strangness, Spectra, HBT) do not analyze directly STAR DSTs (primarily for speed reasons), using instead ntuple/tree-based DSTs with similar event/track cuts Flow pDST event info (incl RP ) tracks (usually primary) passing qual. cuts Strange DST event info V0, kink, , candiates (w/ info re: daughters) HBT DST ~ 23 x smaller than STAR DST (event.root files w/no hits) require 9TB/23 ~ 400 GB for single QM02 production “big four” could almost keep their QM02 DSTs on 2x800 GB PWG disks (but would need some more, to keep 130 GeV data as well) But again, let’s say yes, barely, the big four could continue as usual
ma lisa - STAR collab mtg feb20026 But… should we continue this way? Yes it’s easier-- we are all set up and used to it (and don’t bother me when QM02 is on the horizon) No it’s a bit silly to write essentially the same information into several files even the big four’s DSTs will grow when FTPC information is included the big four have new friends who might like some storage space too TOFp, RICH, high-p T (??) Spin, EMC all have their own DSTs now fragmenting the data (tracks in one ntuple, EMC in another, V0s elsewhere) is short-sighted and does not exploit full potential of STAR e.g. v 2 (V0) analysis could not use flow pDST and resorted to (slow) DSTs (well, could have used HBT DST…) IMHO: even if we get away with it this time, at some point in (near) future, we will have to come to same solution to which every mature analysis/experiment comes -- we should start and get experience with this common datasource encourages common cut values (centrality…)
ma lisa - STAR collab mtg feb20027 But… are we ready for a common DST? (common = all detectors in + all groups could use) Well, yes, as long as we are not too dogmatic… Big Four: settled on contents & quality cuts for TPC info ~ high compression newer systems still developing cuts/analyses would prefer less compression EMC depending on analysis (pp or AA), need hits, clusters, tower sums keeping all (redundant info) means data expansion ~x2 relative raw data ~15KB/evt central AuAu (raw data ~8KB/evt) on-the-fly expansion ~ 1 sec/evt RICH and TOFp compression could be quite high eventually, but for now… each would desire ~ 1KB/evt (central AuAu) FPD ?tiny? ?include? FTPC assume similar information volume/track as main TPC
ma lisa - STAR collab mtg feb20028 A possible “common” DST Sum of:kB/evt (central AuAu) present HBT DST (tracks, V0s kinks…) 80 EMC expanded info 15 (when present!) TOFp 1 RICH 1 FTPC “pessimistic” (?)100 Total ~200 This still represents a compression x 1/9 relative to STAR DST, (and it would be faster) This would reside on production space (11 TB) not PWG space (DSTs not disk-resident) space for multiple versions (think ahead…)
ma lisa - STAR collab mtg feb20029 Specific proposal Adopt common DST, including “fat” data from new subsystems implementation “volunteers”: Frank & Mike In production, create event.root files, “all” of which migrate quickly to HPSS common DST, which remain disk-resident in production area (freeing space on PWG disks) in principle could also create special DSTs for RICH, TOFp, EMC ~10% “extra charge” subsystems should act “responsibly” towards common DST (migrate improvements) common DST must be re-generable from event.root files proper Maker taking StEvent as input can fix small calib etc. issues without full reproduction
ma lisa - STAR collab mtg feb Summary of I Straw-man QM02 sets scale for… time requirements we can meet straw-man physics goals with some safety, but we must be serious about readiness/homogeneity storage requirements we could continue as usual, but would probably regret it - see previous reasons STAR-wide common DST proposed HBT- DST + “fat” data from new systems speed, storage, commonality benefits volunteers for implementation if folks are on-board
ma lisa - STAR collab mtg feb Strawman QM02 productionplan - iteration Au200Au central fullfield3.0 Mevts7.9 wk yield & crude spectrum high p T spectrum to ~10 GeV -bar K 0 -K 0 HBT Au200Au minbias fullfield3.0 Mevts2.2 wk v 2 of K 0, high p T v 2 Au200Au central halffield0.25 Mevts0.65 wk HBT: R (p T ), K- correlations low p T spectra Au200Au minbias halffield0.55 Mevts0.55 wk HBT: R (p T, ) pp triggers (incl hightower)?10.0 Mevts?2.3 wk spin analyses high pT reference K- correlations comments?
ma lisa - STAR collab mtg feb Ordering of data processed… some issues In general, process data from end of run backwards Jerome willing to run 2 datasets (e.g. AuAu-central, pp) parallel Hold off on post-13 Nov due to calibration issues? high p T, spin groups want large-statistics end-of-run pp “early” SVT group wants this dataset run “late” tentative compromise target-date: 15 March FTPC - expected reco upgrades ready mid-March - need short test-production and then likely reproduction ~ 80Kevts - “little” things like this should not perturb us too much, but we should try hard to maintain homogeneity of QM02 production