Trigger Table Improvements past present and future prospects
Trigger Table Improvements -- past, present and future prospects Kirsten & Ted, Feb. 02, 2006, CDF meeting Happy Chinese/Korea New Year! Year of dog… When it comes to trigger, stay low (in rate) !. Photo taken by Ted: Party @Y 2 K’s mansion, Jan. 29
One year of bandwidth improvements L 1 A bandwidth improvements (many heros) L 2 Pulsar upgrade, followed by Pulsar. II with Filar (from March, 2005) SVT AMSRW/HF++ upgrade (Summer 2005) improved L 1 A bandwidth above 30 KHz SVT HB++ is almost there (more improvement!) Silicon L 1 A limit increased from 30 KHz to 40 KHz (Sebastian) (Nov. 2005) L 2 A bandwidth improvements (Markus !) EVB upgrade, improved from ~370 Hz to ~800! (Aug. 2005) L 3 upgrade, more than double the CPU processing power! (on going!) XFT upgrade is on going! (crucial to get us to 300 E 30) CSL/Raw data logger upgrade is coming… overall throughput to tape from current ~20 MB/s to ~80 MB/s “Almost to the impossibly ambitious original L 1 A spec, and well past the original L 2/L 3 rates”. -- Jim Patrick’s comment on Jan 13, 2006. Always keep in mind!: Real goal is to improve bandwidth at the “system level”: from Collision point --> Detector/L 1/L 2/L 3/csl/tapes/disk/ntuples/users/mtgs/GP…. --> Phys. Rev. editor’s desk!
One year of Trigger Table Performance Improvements Timeline Feb. 2005: Mar. 2005: April 2005: July 2005: Aug. 2005: Jan. 2006: 3_00_v 3 (default & highlumi) 3_01_v 1 (version for L 2 Pulsar driving) 3_02_v 1 (1 st attempt at ROF at L 2) 3_04_v 1 (really removed ROF at L 2) 3_05_v 1 (XTRP firmware change, Mt) 3_06_v 1 (A Major clean up of tables) 3_07_v 1 (1 st attempt at unification) 3_09_v 1 (1 st GUT table in action) For details, see TDWG meeting talks between June-Sept 2005, and Jan. 2006.
The goal of GUT table is to get rid of HIGHLUMI table, to allow most triggers run into higher luminosity… The following datasets are not in the highlumi table, and now they are running at higher luminosity in the GUT table… B_BACKUP B_ELECTRON B_HADRONIC B_MUON B_PIPI DIFFRACTIVE HIGGS_MULTI_JET HIGH_PT_BJET JET_20_CENTRAL JPSIEE Z_BB About ~ 60 triggers at L 2
One year (2005) of Trigger Table Performance with beam limited by L 2 A 1 st GUT (Grand Unification Table) 3_09 3_07 3_06 3_02 limited by L 1 A 1 st attempt at unification limited by L 1 A Major clean up default Major clean up ~10% DT, limited by L 1 A highlumi 90 100 130 170 Luminosity (E 30) 200
Initial look at GUT 3_09_v 1 performance: L 1 & deadtime Scen. A_UPS enabled Scen. C_CLCM enabled Z_to_BBbar enabled Default 3_07 HIGHLUMI Take advantage of: • Silicon L 1 A limit increased by Sebastian • L 2/SVT/EVB/L 3 improvements Details see: TDWG web talks on Jan 06 & Jan. 20 th High luminsity store 4590 Taken Jan. 13, 2006
Initial look at GUT 3_09_v 1 performance: L 2 As expected, GUT 3_09_v 1 performance is limited by initial L 2 A rate, ~600 Hz @165 E 30… The “wall” is around ~700 -800 Hz at higher luminosity…. Need to control the initial L 2 A rate !
rate x-sec MET 25+2 JETs L 2_CMX 6_PT 15 Some good triggers “gone wild” at L 2
Initial look at GUT 3_09_v 1 performance: L 3&CSL output We are now writing more data to tape… E-log entry: -- Fri Jan 13 17: 52: 15: The CSL rate to disk is higher than previously seen and the tape writing rate is ~ 8 -16 MB/s (24 MB/s maximum rate). It looks as though we can likely write data to tape fast enough for now, but it will be close. Doug Benjamin and Krzysztof Genser This is a good problem to have ….
We have learned a great deal from the GUT 3_09_v 1 Run 210008 The GUT table performance will be limited by initial max L 2 A (~800 Hz) Just because your triggers are in Run. IIb list, doesn’t mean you can relax! Some of them could “gone wild” at higher luminosity … The extrapolations in your study could have limitations … Some backup triggers could “gone really wild” as well ! Think harder and study harder during shutdown Hope/beg for another new record store before shutdown to learn more! For details, see TDWG meeting web talk on Jan 20, 2006, or directly: http: //hep. uchicago. edu/~thliu/projects/TDWG_Jan 20_06. pdf
Run. IIb Physics Priorities and Trigger Committee CDFnote 8055 Lots of good work done by many (led by Luciano) ! TDWG takes the recommendations very seriously
Run. IIb Physics Priorities and Trigger Committee -- “zoom in” view (any familiar faces? ) 2 b or not 2 b? no longer the question! Essences of Run. IIb Report: • A list of “Most Important Triggers (MIT)” • Specific recommendations on threshold changes etc for some MITs Run. IIb officially starts from this report !
Is Run. IIb for b physics too? -- B group’s trigger plan B group is making rapid progress on finalizing their plan (not covered in the Run. IIb report) --> Led by Kevin CDF note is in progress… Details see Kevin’s talk at TDWG meeting on Jan. 20, 2006 http: //www-cdf. fnal. gov/internal/Web. Talks/Archive/060120_tdwg/ 04_060120_tdwg_Kevin_1_B_plan_tdwg. pdf Will try our best to implement some of their plan before shutdown….
Goal Setting Spokes ask us to jump, we say “how high? ”… The goal is to reach 300 E 30 with one table, while optimizing the middle&low ranges … Have done reasonably well in the past year towards this goal … 300 E 30 Still a long way to go … TDWG
How to get there from here? -- learn from the past Table clean up: need to do more “Prescale away” more&more triggers (not in Run. IIb list) at higher&higher luminosity … need to do more Improve trigger purity: need to do MUCH more! Optimize middle&low ranges: to do more (B group’s plan) Take crucial data for trigger improvement study: on going Run. IIb threshold changes: on going! XFT upgrade: eager to see how much improvements ! Implement a few good new triggers along the way … Learn as much as we can before shutdown … on going Brain-storm on new ideas during shutdown (have a workshop? ) (e. g. see the end of Alberto’s talk)…
One comment about table cleanup Last Aug. we did our first major table clean up, in Table 3_06, by removing junk triggers or clearly unused triggers. that helped us a lot! For details see: http: //hep. uchicago. edu/~thliu/projects/TDWG_July 0805. pdf We still have >~ 130 L 2 triggers, are ALL of them actually really used? Estimate: ~ 100 physicists actively doing analysis (or running CAF jobs? ), many of them use the same triggers…. Backup triggers could cause trouble at high luminosity… and there are lots of them. Veronica will set up a web page with the list of the triggers in the current table, please take a look and “claim” the triggers you are really using… We will do this during the shutdown, and any unclaimed triggers will be dropped later.
A slide from last TDWG meeting (July 08), on Table cleaning Our table is 5 years old, designed > 7 years ago. Has been used to serve food for > 600 physicists (lots of them are poor, hungry and hardworking young postdoc and students). Over years, it has accumulated dust, food stain … it has been burned by high luminosity beam quite many times and broken into two pieces… “High Lumi Table” == “Panic Table” It is time to do a clean up, so the young people can eat healthy/clean food….
It has been a real team effort: Physics+Hardware+Software+Ops TDWG Heads: Kirsten and Ted Trigger Table Creation and Testing Coordinator: Veronica Sorin / Zeynep Gunay Technical Support: Trigger SPLs Vadim Rusu/Greg Feild/Veronique Boisvert Trigger Software Coordinator: Daniel Whiteson Trigger Database: Tom Wright / Donatella Torretta XMON: Charles Plager Retired Generals: Kevin and Jonathan Run. IIb Committee: chaired by Luciano Trigger Representatives: Exotics Group: Vadim Rusu and Oscar Gonzalez Lopez QCD Group: Mary Convery, Gene Flanagan and Mario Campanelli Top Group: Veronique Boisvert and Ambra Gresele EWK Group: Jiyeon Han B Physics: Konstantin Anikeev and Simone Donati (with Rolf Oldeman) Hardware teams: Sili SPLs (Sebastian/Ankush/Marcel), SVT team(Alberto etc), EVB/L 3 team (Markus), XTRP (Chris Marino) … Others: Chris Neu, Andrew Ivanov, Viktor Veszpremi, Bill Badgett, Peter Wilson, Carla Pilcher, Monica Tecchio, Camille Ginsburg … Thanks to all physics conveners, and ops managers.
How long it may take to reach 300 E 30? -- hard to predict, but can use data points from last year… Months taken to improve table vs its luminosity range ? 12 3_09 9 6 ~ 1 year from now Matches well with machine schedule 3_07 3 0 3_02 100 E 30 200 E 30 300 E 30
Will there be an end? 3_09: GUT table Ultimate ideal goal: TOE (Table Of Everything) Is it possible? Probably not! But who could stop us from trying hard ? … at least @ the year of dog… (next year is pig…. so there could be an end after all !)
- Slides: 20