GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
The Plan
• The Year in Review
• The Big Picture
• Short Term Plan & Balloon
• Manpower issues
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
The Year in Review (1)
•
Just over a year on the job for me
– Initial conditions
• AO response in prep
• TB99 underway
•
Early decisions
– Time to address GlastSim
infrastructure first
•
•
•
•
Code build & management
Architecture
I/O
Flexible geometry
– Move to GEANT4
• Try first on balloon
– Support 3 operating sys only
• WindowsNT, Linux, Sun
– Try out Root in TB99
– Science Analysis S/W subgroup
•
Manpower Issues
– Jose Hernandez, Sawyer
Gillespie, Jeff Silvis move on
• No TKR expertise
• ACD hit
• Core group down to 2.5 FTE
– Did manpower projection
• Need 10-12 FTEs for core
development
• Need new TKR folks
– Approval to hire SciProg at
SLAC
• Karl Young, starts 2/1/2001
– Hired 2nd UVic CoOp student
• Ian Gable, started 9/2000
– Worked on expanding effort
to Italy and Japan
• Riccardo Giannitrapani &
Masanobu Ozaki
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
The Year in Review (2)
•
Decisions Implemented
– Break up code into packages
– Separate Sim + Recon
– CMT for code management and
build
– GAUDI for architecture
– See core review for details on
Gaudi & CMT
– Root I/O used for TB99
• Root
IDL converter
• Direct Root analysis
– G4 in use for Balloon
– Big effort on documentation
– Moved cvs repository to SLAC
• Industrial strength server
• Ditto for software web
•
Consequences
– GlastSim has been “in the
shop” for a year…
– TB99, turnover and small
core group resulted in slow
progress
• CMT ready; cvs repository
moved to SLAC in Summer
• Gaudi work started for real in
Fall
– Made it difficult to try new
things for GTOCC, CAL and
ACD studies
• Used AO version of GlastSim
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
The Year in Review (3)
•
•
Science Analysis
– Led by Seth & Jay
– Lively sessions at all the
workshops
– Identified big ticket tools that
need to be implemented and
rough idea of manpower
req’d
– Ready to start apportioning
initial tasks
TKR support
– Tracy Usher and Leon
Rochester on board and
coming up to speed
– Expect to have strong group
– SLAC, UCSC, Italy
– See TKR review
•
Communications/Groupware
– Software workshops
quarterly
• 2 Subsystems reviewed each
time
– Weekly software meetings on
vrvs
• CAL and Core now have their
own weekly vrvs meetings
– ICQ has pockets of popularity
• Those who use it really like it
• Windows + Linux
– MS Project schedule and
Workplan document
developed, maintained
http://www-glast.slac.stanford.edu/Software/
• Available from software
home page
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
The Big Picture
Orphaned
Addressed in 2001 timeframe
Sources Simulation
Science
Tools
Sim+Recon
Trigger Analysis
PSF/Aeff optimization
Data/MC
Production
Facility
Background Rejection
Level 1
Analysis
Platforms
Calibrations
Event
Display
Release
Management &
Validation
Performance
Monitoring
Database
Development
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Timescales
What
Start/Due
Responsible
Sources
ongoing
Hiroshima + TBD
Sim+Recon
yesterday
Core+Subsystems
Trigger Analysis
Start mid 2001?
PSF/Aeff
ongoing
Bkg rejection
ongoing
Calibrations
Start mid 2001
Eduardo
Mark I Due early
2002
SLAC
Data Prod Facility
Description
Auto-server, RDB
Event Display
now
Release Man/Verify
Due mid 2001
Karl Young
Performance Monitor
Due end 2001
Steve
Database Devel
Start mid 2002
Science Tools
Start FY2002
Analysis Platforms
Ongoing/balloon
Seth & Jay
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Timeline
2001
2002
2003
2004
2005
2006
Balloon
GlastSim/G4
Event Display
Release Manage
Performance Tuning,
Tracking
Data Production Facility
+ monitoring
Calibrations
Science Tools, DB, Analysis Platforms
Polishing & maintenance + post-launch panic
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Science Data Production - Picture
Ground
Station
WWW - code, algorithms
MOC
mirror
PACOR
in box
IOC
IOC
Ops
HSM
Level 0
Proc
Server
SSC
High-level
monitoring
HEASARC
Photon DB Photon DB
RDB
Event DB
Recon
Level 2
Level 1
MC
Calibrations
SSC
Access to Data
PI Teams
HSM = hierarchical storage manager
RDB = relational database
Guest Obs
IOC
Access to Data
SOC
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
From Space to Photons
G4/Gismo
MC-Digi
MC Hits
Raw tele
Level 0
IOC
“raw” data
Raw-Digi
Phys
ACD, CAL, TKR
digitizations
Phys, Calib
ConstantsDB
Calib
Trigger
HK
PreProcess
“Eng” units
Subsys independent
up to here
Livetime
Elec Calibs
Eng data
ConstantsDB
Deal w/ HK
Constants DB
ConstantsDB
CAL Rec
CAL Recon
clusters
Event
Bkg Rej
TKR Recon
GCR Calibs
tracks
TKR Rec
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Short Term Plan (1)
•
•
See Core, TKR reviews, CAL &
ACD status talks for activities
– CAL has program of
sim/recon upgrades for 2001
Core effort is to move GlastSim
into the new architecture in time
for the balloon
– Incrementally approach final
product via increasingly
complete prototypes
– Started from repackage of
Recon with IRF as input
• CAL, TKR now functional in
Gaudi framework
• We have the technology
– Next steps are to replace IRF
and break out separate digi
phase
•
Have defined MC Truth classes
– Divide volumes into “step” and
“integrating” types
• “Step”s record (x,p) details of
each particle’s passage
• “Integrating”s sum up energy
loss, keeping list of dE per
‘primary’ particle
– Position info kept as energyweighted moments
– Masanobu implementing in G4
•
Digitization classes now under active
discussion
– Based on TB99 versions
– ACD approved
• Heather producing first full-chain
prototype
– TKR approved
– CAL still under discussion
– Will just take existing digi algs as
first attempt
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Short Term Plan (2)
•
•
Next steps are to modify recons
to eat digitization classes and
output recon classes
– Replace ntuple
– Start from TB99 recon
classes
– Direct access to all Track &
Cluster info in output files
– Re-enable event display
Critical Path
– MC Truth implemented in
simulator (G4, Gismo)
– Digis extracted from
GlastSim
– Recon eats digis
– Schedule says end of Jan
•
Geometry is main Caveat
– Separated Sim + recon broke
existing geometry model
• kludged
– GlastSim scheme required
sim + recon to live together
to give access to the
geometry objects
• We want a portable geometry
– Also want flexibility
• Ability to ‘discover’ the
desired geometry from the
input xml ascii description
• Build the geometry that was
found, without coded
assumptions
– Joanne is working on an ACD
prototype
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Short Term Plan (3)
•
Balloon
– Now have raw format from JJ
et al
• Heather will write converter
to digi classes
– Once prototype is together
• Time to hand over
subsystem code
• Start testing with G4 and
hopefully Gismo
– More in Balloon session
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Short Term MS Project
To be filled in closer to time of talk
with latest version
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Organization
Code Architect:
T.Burnett (UW)
Sim/Recon:
T.Burnett (UW)
Manager:
R.Dubois (SLAC)
Calibrations:
E. Do Couto e Silva
(SLAC)
TKR:
Tracy Usher (SLAC)
CAL:
E.Grove (NRL)
A.Djannati-Atai (France)
ACD:
H.Kelly (GSFC)
Science Analysis:
J.Norris,
S.Digel (GSFC)
Analysis Tools:
H.Kelly (GSFC)
Infrastructure:
Karl Young (SLAC)
• use cvs for distributed code development
• use web conferencing tools for meetings
• use instant messager tool for quicky discussions
• software team is integrated
• subsystem folks are matrixed between subsystem
and software group
• part of same development process as other
components of software
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Manpower Issues
•
•
There are always manpower
issues!
I still recall Jonathan Ormes’
statement a year ago
– “We low-balled software by a
factor of 3 in the AO
Response”
– OK, so what do we do about
it?
•
Karl Young starts in 2 weeks
– Should make a big difference
to core team
•
We are hoping to get help from
Italy on
– TKR simulation
– GEANT4 czar
– GEANT4 geometry
– Event display
•
We are hoping to get help from
Hiroshima on
– Sources
Recruit!
•
We are presuming Toby will get
his NSF grant
– SciProg + grad student in mid
2001?
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Outlook
•
We are starting to cook with gas!
•
BUT we are still left with big open
areas
•
– Performance optimization
and monitoring
– Background rejection
– Trigger analysis
– Root expertise
•
•
The plan is to have the new
GlastSim “complete” by Fall 2001
Then start shifting attention
downstream
– Calibration
– Science/analysis tools
– DPF
I still don’t know how the Data
Production Facility gets
developed…
R.Dubois
GLAST LAT Offline Software
Workshop - SLAC, Jan. 16-19, 2001
Workshop Goals
• Bring people up to date on code migrations
– TKR, Core reviews, status talks
• Balloon software status and directions
• Science analysis
– Start getting down to business
• Calibrations
– Construction databases
– Start organizing subsystems
• Bring Italian group on board
• Group input on directions
R.Dubois