25
eter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Embed Size (px)

Citation preview

Page 1: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

State of the Software

Peter SteinbergSoftware President

January 24, 2002

Page 2: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Outline

• Data handling• DAQ / GUI• Distributor/Mover• Data Validation• Event Display

• Detectors• Plastics/Trigger• Silicon• PCAL• SpecTrig

• Infrastructure• Geometry• DB Access • Phat][• AnT Trees• File catalog / DD

• Analysis code• Event Selection• Data Tracking• Tracking• Vertexing

Page 3: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Data Handling

Page 4: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

DAQ & DAQ GUI• Andrei & Pradeep• GUI (Pradeep)

• Shift mode as of Wednesday• Documentation coming

• Functionality• silicon, latchup, rocdb, FBTIME/PED

• Distributor & DAQ & ROCDB play nice• Crash recovery• Todo

• Trigger configurations not yet downloaded into L0/L1 managers

• Need Trigger config synced w/ DB

Page 5: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Distributor & Data Mover, etc.• Burt• Distributor in stable state

• DAQ work has improved things

• To do• Hunt down remaining segvios• Hanging-socket problem in DAQ• Document & bullet proof client that runs modules

(needed for online processing)

• Data Mover• X2 rate by concurrent checksums, x2 from

multithreading expected

• Pedestals• Datamover now starts ped running

Page 6: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Data Validation

• Data Validation• 8 hour jobs with 1 GB files• Do we plan to increase this?• Intense need for profiling and optimization

• Compacting files• When do we write all collisions?• Zero Suppression• Writing out Object Managers every 10 files

• MBT Monitoring instabilities in DV, HPSS• Reference histograms?• Consolidate output files: AnT Maker?

Page 7: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Event Display

• “phever” – online trigger event display

• Currently getting “raw” events from the distributor• Separate trigger processing envisioned• Better response from phever

• To do• Incorporate expert plots• Get alarm requests• Compiled version• Printable summary plot if needed

Page 8: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Detectors

Page 9: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Trigger

• Dave et al• Online client needs to be tuned for this

year’s data• Lots of items integrated into event display

• Aneta writing TrgInspector• Will keep processing TrgTrees and

InfoTrees to monitor health of trigger

• Trigger info on the web• Trigger info in the data

• TrgEventData still being updated• TrgTrees have all the new detectors

Page 10: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Plastics

• Frank• Calibrations

• New triggering makes old TOF calibration procedures suboptimal

• FB works fine

• Calibrations that need DV output are highest priority to fix

• Online clients will come next• What about catching bad TDCs?• Inkyu’s client is useful, but needs more

interaction to determine optimal settings

Page 11: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

PCAL• Corey & GSFS• Tasks finished

• VME unpacking, PCAL Event data, TrgTrees, Calibrations, event display

• Todo• Online monitoring clients• Online pedestals • Raw branch in TrgTree

• Longer term• PCAL gain calibrations• Geometry in PMC• Full MC implementation

Page 12: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

SpecTrig

• Gabor• Done

• Online client running – no monitoring• Some detector checks, not automated• Trigger bits are in data (TDC & Latch)

• Todo• Data tracking – collaborate w/ TrgGroup• Dealing with trigger mixes• VME data in event, trees

• Longer term• Time calibrations?

Page 13: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Silicon• Gerrit, Rachid, Christof• Online latchup clients still running – no

changes expected• Pedestals hard-wired to run on sequence 0.

Gerrit will deal w/ reference sets• No more calibration runs (DAQ issue)

• Running with master gain file from last year• Not automatic – just “replacing” file as needed

• Running fine in DV – DCM main issue• Will continue to not apply it to final hit arrays• Some controversy about access?

• Silicon processing not finished• Rings still done the old way (Yuriy)

Page 14: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Analysis Infrastructure

Page 15: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Geometry

• Adam (geometry), Andrzej (access)

• Big change: geometry from DB• Unfortunately, tripped off bug in phat• Back to file access

• LatestGeometry will be retired• Can get “latest” by PR0x run

• Some issues with dealing with frames and supports of new detectors

Page 16: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

DB Access

• Still GAH’s code, KG & BW • Serious issue with dangling cursors in Oracle• Kris & Bolek working w/ Marguerite to debug• GAH put in “auto-connecting” in DB software

– not good behavior• KG proposes to remove, but this will wait until

things stabilize

• Kris will put up web page explaining how best to use TPhDatabase

• No comment yet on DbKeys…

Page 17: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Phat][

• Maarten & Peter

• Slowed due to work on the run

• Finishing up retooling of modulelib

• Macros will be imported at a minimal level – if you need it, import it!• Old macros will obviously be in Phat6 CVS

• Volunteers coming on shell

• Trivial to check out of CVS

Page 18: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

AnT Trees

• Maarten & JWL

• This will be the main mini-DST format• Trigger information for event selection• Hits & Tracks• No multiplicity details planned

• Need requests for more functionality• (me) Make it modular so it’s easy to add

objects as branches (a la EventInfo)

• Integrate more of the needed data

• Add to production

Page 19: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Data Accessibility

• Maarten & Jinlong• CatWeb

• Managing 130,000 files, 175 TB• 3 disk pools (RCF, Pharm, PDEV)• 50k file instances (16 TB)• 300,000 accesses of file instances via TPhDST• 2-5% staging error rate (HPSS, RCF, user

problems)

• Plans• Filesets, gui improvements, improve job

scheduling, disk cleanup features• Have phat speak to CatWeb server, not Oracle

Page 20: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Analysis

Page 21: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Data Tracking

• Dave et al

• No consistent data tracking scheme created yet

• Pradeep will update scripts doing run totals online• PAS has already written scripts using the

DAQ-generated run totals (vs. B)

• Scalers in the data stream• Need verification and matching to visual

scalers

Page 22: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Event Selection

• Dave et al

• Done by macros at the moment• Full IsCollision(), suitable for DV, coming

early next week

• Centrality cuts released to collaboration• First round from richard• More to come

• Not clear how to handle trigger mix• Major issue for recombining different

hardware data selections

Page 23: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Tracking

• Conor & Abby, DV by Jang-Woo

• New supermodule planned• PR03LowMultTracking

• Major changes• Vertex-less straight tracking• Then vertex from straight tracks

• Changes related to new field map

Page 24: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Vertexing

• Andrzej & Krszystof• All vertexing procedures work

• Single track vertexing not yet in phat, but clearly crucial for d+Au and p+p run

• OctDeVertex: 90% eff, 1-1.5cm res• OctagonVertex: 80% eff, 2.5cm res• Zvertex: 35% eff, best y• SpecMainVertex: 10%

• Forcing OctagonVertex to agree w/ OctDeVertex actually reduces purity of other vertices

Page 25: Peter Steinberg State of the Software Peter Steinberg Software President January 24, 2002

Peter Steinberg

Conclusions• Data Handling

• Trigger configuration in DAQ• Data file sizes / production time

• Detectors• Monitoring, monitoring, monitoring • Si silicon processing (e.g. rings) not complete• TOF calibration infrastructure crucial

• Analysis Infrastructure• Phat][ imminent• DB stability potential production problem

• Analysis Software• Event selection, vertexing, tracking all coming on

shell