P116 Reprocessing

status: Under Construction Complete
last update: 14 October 30 November 2010

This page is a record of the configuration for the P116 reprocessing project, rebuilding FT1 files (and friends) for all of science data with new, extended event classifications.

• P116-FT1 - this task will read existing (P105+level1Level1) MERIT files and produce FT1 (photons) + LS1 ("fat" FT1) files

...

/glast/Data/Flight/Reprocess/<reprocessName>/<dataType>

Locations for P120P116:

/glast/Data/Flight/Reprocess/P116/ft1
/glast/Data/Flight/Reprocess/P116/electronft1
/glast/Data/Flight/Reprocess/P116/ls1

...

Note: 'procVer' is a field added to the file name (and the keyword "PROC_VER" in the primary header) added to the FFD 5/12/2010. Ref: http://fermi.gsfc.nasa.gov/ssc/dev/current_documents/Science_DP_ICD_RevA.pdf^{Image Removed}

Example:

/glast/Data/Flight/Reprocess/P120P116/ft1/gll_ph_p116_r0239559565_v116.fit
/glast/Data/Flight/Reprocess/P120P116/ls1/gll_ev_p116_r0239559565_v116.fit

...

Data/Flight/Reprocess/P116:FT1 r0239557414
Data/Flight/Reprocess/P116:LS1 r0239557414

Data Sample

The currently defined data sample for P116 reprocessing includes :the following runs.

Note that the merit files used for input to this reprocessing derive from two sources:

Final number of selected photon events (in the FT1 and LS1 files) = 435,801,176

These data are destined for the FSSC and were made public as of 17 November 2010
These data (along with subsequent Level 1 data) are in the SLAC Astroserver as P6_public_v2.

...

Three extra runs were unintentionally processed as part of P116:

The first of these was deemed good for science, while the last two were not. Therefore, runs 250687192 and 250692922 have been retroactively removed from the datacatlog, astroserver and FSSC. (Removed from dataCatalog on 2/22/2011.)It is expected that a backfill operation will happen to mesh with Level 1 once it begins to produce these new event classifications. Ultimately, these data are destined for the FSSC.

Bookkeeping

1. (This page): Define ingredients of reprocessing (processing code/configuration changes)
2. Processing History database: http://glast-ground.slac.stanford.edu/HistoryProcessing/HProcessingRuns.jsp?processingname=P116^{Image Removed}
   1. List of all reprocessings
   2. List of all data runs reprocessed
   3. Pointers to all input data files (-> dataCatalog)
   4. Pointers to associated task processes (-> Pipeline II status)
3. Data Catalog database: http://glast-ground.slac.stanford.edu/DataCatalog/folder.jsp^{Image Removed}
   1. Lists of and pointers to all output data files
   2. Meta data associated with each output data product

P116-FT1

Anchor
P116-FT1 P116-FT1

(This task is roughly equivalent to the P105-FT1 task but with updates and modernization.)

Status chronology

• 11/29/2010 - Run 245403855 start time modified in datacatalog (1 second earlier), regenerate runFile.txt, and rollback stream 1019 to make astroserver happy. This adds three (3) events to that run. (Statistics above not updated)
• 11/17/2010 - reprocess final backfill block3 (98 runs)
• 11/12/2010 - discover that gtdiffrsp params were incorrect (evtclass parm ignored, so all events got all diff rsp calc). Next (last) backfill block will be corrected.
• 11/11/2010 - gear up for reprocessing data block2 (617 runs)
• 10/18/2010 - 146 mergeClumps jobs ran out of CPU time. Changed to xxl batch queue and rolled back.10/1/2010 - Begin task construction
• 10/14/2010 - Validation complete, begin production for data block1
• 10/1/2010 - Begin task construction

Configuration

Task Location	/nfs/farm/g/glast/u38/Reprocess-tasks/P116-FT1
Task Status	http://glast-ground.slac.stanford.edu/Pipeline-II/index.jspImage Removed P116-FT1
Input Data Selection	MERIT
Input Run List	ftphttp://ftpglast-glastground.slac.stanford.edu/Decorator/Decorate/glast.u38/Reprocess-tasks/P116-FT1/config/runFile.txtImage Removed
Reprocessing Mode	reFT1
evtClassDefs	00-18-00
photon cut	pass6_FSW_cuts
eventClassifier	dataclean_classifier.py
eventClassMap	n/a
s/c data	FT2 from P105 (runs 239557414 - 271844560), then from current Level 1 production
ScienceTools	09-18-04 05 (SCons build)
Allowed Code Variants	redhat4-i686-32bit-gcc34, redhat4-x86_64-64bit-gcc34, redhat5-i686-32bit-gcc41, redhat5-x86_64-64bit-gcc41 (Optimized)
Diffuse Model	/afs/slac.stanford.edu/g/glast/ground/releases/analysisFiles/diffuse/v2/source_model_v02.xml ( https://confluence.slac.stanford.edu/display/SCIGRPS/Diffuse+Model+for+Analysis+of+LAT+DataImage Removed )
Diffuse Response IRFs	P6_V3_DIFFUSE, P6_V3_DATACLEAN
Output Data Products	FT1, LS1

...

Note on 'Code Variant': The SLAC batch farm contains a mixture of architectures , both hardware (Intel/AMD 64-bit) and software (RedHat Enterprise Linux 4, and 5, gcc 3.4, 4.1, etc.). Currently, all batch machines have 64-bit hardware and (most, if not all) run with the 64-bit operating systems.

Note on diffuse response calculation: gtdiffrsp is called two times in succession. The first time with IRF P7_V2For the bulk of this processing, the incorrect parameter was used for gtdiffrsp (IRF P6_V3_DIFFUSE and evclass==3, followed by and IRF P7P6_32V3_dataclean and evclass==4). After block2 this was corrected to IRF P6_V3_DIFFUSE and evclsmin==3, and IRF P6_V3_DATACLEAN and evclsmin==4. The resulting FT1 file has four columns of diffuse response, two columns (galactic and extragalactic response) for each of the two IRFs.

...

files have five columns for diffuse response of which two pairs are filled in for DATACLEAN, one pair for DIFFUSE and zero for all other events. Each pair contains galactic and extragalactic response.

Timing

• The most time-consuming step in P116-FT1 is the 'mergeClump' step wherein the diffuse response is calculated (twice). The CPU time required for this step, after ~2300 jobs have completed appears in the following plot:
  Image Added
  Thus, depending on the run and the speed of the batch machine (bimodal distribution for fell vs hequ), the jobs are finishing up after anywhere from 2 hours to 11 hours, with the most probable being about 5 hours.

With 1000 cores, one might then estimate 60 hours to complete the entire project. Near final timing
Image Added

After the correct gtdiffrsp parameters were established, the mergeClumps step was reduced to ~30 minutes each.