Trigger Menus and Rates Trigger Menu Development Group: Leonard Apanasevich, Anne Fleur Barfuss, Pedram Bargassa, Eric Conte, Sivia Goy Lopez, Jonathan J. Hollar, Chi Nhan Nguyen CMS Physics Week May 13th, 2009
Introduction Proposed Menus in 2008: One single menu with >160 HLT paths which was supposed to cover a wide luminosity range L = 8.0 1029 1.0 1032 cm 2s 1 Change of Strategy: Decided to reduce number of trigger paths (maintenance, manpower) Different menus for 3 reference luminosities: 8.0 1029 (8E29), 1.0 1031 (1E31) and 1.0 1032 cm 2s 1 (1E32) Initial version of lean 8E29 and 1E31 menus were presented to all groups shortly after November 2008 CMS week in order to be reviewed by all Trigger Reviews: Thu Dec 11 (CMS week): JetMET Mon Dec 15: Muons Wed Dec 17: Egamma Wed Jan 14: b tag/tau/p flow Wed Jan 21: Cross triggers Wed Feb 4: AlCa/MinBias/Commissioning triggers After reviews and following approval process: Trigger Development Group (TMD) started to compile the menus and evaluate the rates 2
Software & Samples Many new trigger paths not yet implemented in 22X release: Running OpenHLT on Summer08 samples, considered as best choice to give estimates on rates and give fast feedback to groups OpenHLT has been validated several times and has been shown that it can consistently reproduce rates made from regular HLT software However, some few algorithms cannot be simulated: other estimate approaches applied or planned to Menus & Rates are preliminary! Samples: sbqcd (9.9 M events): /MinBias/Summer08_STARTUP_V5_STARTUP_V5_v1/GEN SIM RAW CMSSW_2_2_5 STARTUP_V8 GlobalTag DigiToRaw redone 8E29: No L1 jet corrections except for flat 0.7 HF correction, also no HLT jet corrections L1 scales optimised for low energies (convertion scheme from ET to GT bit configuration): L1Menu_Commissioning2009_v0_L1T_Scales_20080926_startup_Imp0_Unprescaled_cff 1E31: L1/HLT Jet corrections applied L1 scales optimised for higher energies: L1Menu_MC2009_v0_L1T_Scales_20080922_Imp0_Unprescaled_cff 3
General Remarks Latest versions of menus and more detailed info about definitions, known issues, can be found here (+ links from there): https://twiki.cern.ch/twiki/bin/view/cms/tsg_03_v_09_8e29 https://twiki.cern.ch/twiki/bin/view/cms/tsg_03_v_09_1e31 https://twiki.cern.ch/twiki/bin/view/cms/triggerreviews_0809 General constraints: L1 rate: 17 khz (50 khz / safety factor) HLT rate: 1 GB/s / event size / safety factor 150 Hz for physics triggers AlCa: higher rates at lower event sizes Timing NOT covered here Menus & Rates are preliminary! 4
Menu 8E29 can be considered as a candidate menu to be used for the first real collisions focus on commissioning many triggers are dominated by noise 5
JetMET Triggers (8E29) U indicates uncorrected jets (L1 or HLT) HLT_L1* denotes HLT passtrough triggers (no HLT cut) No known issues 35 (jets) + 5 (MET) Hz Monitoring triggers Physics triggers 6
Muon Triggers (8E29) Loser muon quality cuts for L1_SingleMuOpen than L1_SingleMu0 HLT_L2MuX: Level 2 only (no tracker) reconstruction Iso: L2 calorimetric isolation and L3 tracker isolation applied Note: Effectively all good muons reconstructed at L1+HLT Monitoring triggers will be written to disk Physics triggers + 33 Hz if a physics trigger needs to be prescaled: Backup triggers 7
Electron Triggers (8E29) LW: Large pixel matching window SW: Startup window (narrower than LW) EleId: Electron ID based on cluster shape variables SC10: Additional supercluster with ET>10 GeV required Monitoring triggers No known issues Physics triggers Backup triggers Express Stream + 30 Hz (fast analysis/monitoring on CAF) 8
Photon Triggers (8E29) TrackIso: track isol. (< 2 tracks in iso cone) cannot be simulated in OpenHLT accurately, here rough estimate only LooseEcalIso: < 3 GeV or 10% of photon energy in iso cone rough estimate only Invariant mass window cuts (>2, 2 4.5, 8 11 GeV) taken into account B field bending cannot be done with OpenHLT rough estimate only compare with Roberto's talk from 23 Feb 2009 (http://indico.cern.ch/getfile.py/accesscontribid=10&sessionid=1&resid=1&materialid=slides&confid=51622) + 9 Hz (considered as solid estimation, despite OpenHLT limitations) 9
Tau / b jet Triggers (8E29) U U Taus: No U b/c at HLT correction never applied by default Missing triggers presently not simulated by OpenHLT in this table: HLT_StoppedHSCP + 5 (taus) + 0 (b jets) Hz 10
Cross Triggers (8E29) No known issues + 0 Hz 11
MinBias Triggers (8E29) What is a proper MinBias trigger? Different kinds of possible MB triggers prepared: Hcal based (probably noisy) Ecal based (probably less noisy) Pixel based Track based L1_ZeroBias not available in the 2.2.X: L1_MinBias_HTT10 is used instead + 14 Hz 12
Comissioning Triggers (8E29) HLT_BackwardBSC, HLT_ForwardBSC, HLT_TrackerCosmics: technical L1 bits, not implemented in OpenHLT + 9 Hz (3 missing triggers above will be prescaled to have < 3 Hz) 13
AlCa Triggers (8E29) No known issues Note: AlCa requests more statistics higher rates available at lower event sizes here 14
L1 Menu (8E29) Resulting L1 table: As expected no problems with the bandwidth budget at this luminosity ~5.8 khz 15
Summary: Individual Rates by Group (8E29) about 70 triggers 16
Summary: Cumulative Rates by Group (8E29) Rates by group: 35 + 5 + 34 + 38 + 5 + 0 + 0 + 14 + 9 = 140 Hz Still room for missing HLT_StoppedHSCP and BSC/TrackerCosmics trigger 17
Summary: Data Throughput by Group (8E29) Estimated data throughput of 23 MB/s well within bandwidth budget 18
Menu 1E31 designed for the next large MC exercise 19
Jet Triggers (1E31) Corrections applied for L1 jets and HLT single and double jets Note: here still using obsolete L1 jet correction scheme needs to be re evaluated! DiJetAve paths always use uncorrected jets No known issues 15 Hz 20
MET Triggers (1E31) SumET and HT triggers added HLT_HT250 should have its prescale decreased from 50 to 10 (but not much additional rate expected from that) Missing triggers presently not simulated by OpenHLT HLT_HT200: First tries to simulate it by summing jets, gives 25 (+15 Hz!) Need to check! + 2 Hz (+ 15 Hz!) 21
Muon Triggers (1E31) No known issues + 41 Hz 22
Electron Triggers (1E31) Changed from LW to SW (better alignment expected) Missing triggers presently not simulated by OpenHLT: HLT_Ele15_SW_EleId_L1R, HLT_Ele15_SC15_SW_EleId_L1R, HLT_Ele25_SW_EleId_LooseTrackIso_L1R The following triggers have everything simulated except their special treatment of the B field: HLT_DoubleEle5_SW_Jpsi_L1R HLT_DoubleEle5_SW_Upsilon_L1R + 26 Hz (maybe 5 10 Hz more?) 23
Photon Triggers (1E31) The following triggers cannot be properly simulated with the present version of OpenHLT, so their present rates are just an indication: HLT_Photon10_LooseEcalIso_TrackIso_L1R HLT_Photon20_LooseEcalIso_TrackIso_L1R HLT_Photon25_LooseEcalIso_TrackIso_L1R HLT_DoublePhoton15_VeryLooseEcalIso_L1R + 24 Hz 24
Tau / b jet Triggers (1E31) Changes needed because rate was higher than in review due to bug HLT_SingleIsoTau30_Trk5: L2 Et threshold raised from 20 to 30 GeV HLT_DoubleLooseIsoTau15_Trk5: Additional requirement L3 Tracker isolation is required for at least one of the 2 tau candidates + 14 (taus) + 0 (b jets) Hz 25
Cross Triggers (1E31) Missing triggers presently not simulated by OpenHLT: HLT_Ele10_LW_L1R_HT150 HLT_Mu5_HT50 Not much additional rate expected though + 8 Hz (+ 1 2 Hz?) 26
MinBias Triggers (1E31) No known issues + 9 Hz 27
Comissioning Triggers (1E31) HLT_BackwardBSC, HLT_ForwardBSC, HLT_TrackerCosmics: technical L1 bits, not implemented in OpenHLT + 1 Hz (+ 2 3 Hz?) 28
AlCa Triggers (1E31) HLT_IsoTrack is assigned a rate of 9 Hz "by hand", not yet evaluated. 29
L1 Menu (1E31) Resulting L1 table: Also no problems with the bandwidth budget at this luminosity ~8.6 khz 30
Summary: Individual Rates by Group (1E31) 83 + 6 missing triggers = about 90 triggers 31
Summary: Cumulative Rates by Group (1E31) Rates by group: 15 + 2 + 41 + 50 + 14 + 0 + 8 + 9 + 1 = 142 Hz + 25 30 Hz (?) by adding missing triggers: complete table and prescales to be revisited soon 32
Summary: Data Throughput by Group (1E31) Estimated data throughput of 31 MB/s well within bandwidth budget 33
Summary & Plans Summary: Change of Trigger Menu Strategy From a single menu with wide luminosity coverage to multiple menus for different reference luminosities Preliminary menus & rates for L = 8.0 1029 (8E29), 1.0 1031 (1E31) presented Preliminary rates are within bandwidth limits Though still not all paths can be simulated in OpenHLT All trigger paths are currently being checked one by one New draft versions are being announced regularly Plans: Timing needs to be checked Move to CMSSW 31X to evaluate rates with regular HLT software Starting 1E32 menu (redo HLT exercise) Physics performance (ongoing reviews to validate performance of current menus) 2_2_6_HLT 8E29 menu is currently running in MWGR (first results expected soon) 34