Transcript Software Summary

Software Summary
Database
Data Flow
G4MICE Status & Plans
Detector Reconstruction
M.Ellis - CM24 - 3rd June 2009
1
Database
 Storage of the geometry is progressing well.
 Numerous geometries have been stored to the DB
and then retrieved.
 Work is moving to calibrations and then cabling
information now.
 API exists now at RAL.
 Discussion during the session concluded that the
best solution is for the API to be on the MICE
network (in the control room) and a process can run
on heplnw17 to forward requests from outside of
the lab.
 This way a network failure will not prevent data
taking.
M.Ellis - CM24 - 3rd June 2009
2
API
 Communication with the API is through the use of
XML files.
 G4MICE already has an application that takes the
geometry and generates an XML file.
 The API converts this to SQL (which decouples
G4MICE, the DAQ, and other users from any
changes to the DB including the technology in use).
 Will soon write code to read an XML file into
G4MICE to use a geometry retrieved from the DB.
 Queries to the DB will be possible through a web
interface and through a GUI in the control room.
M.Ellis - CM24 - 3rd June 2009
3
Data Flow
 Long discussion during the plenary presentation
yesterday.
 Key points:

Henry’s talk
http://www.mice.iit.edu/cm/cm24/agenda/cm24_nebrensky_dataflow.ppt

MICE Notes 247 and 253:
http://hep04.phys.iit.edu/cooldemo/micenotes/public/pdf/MICE0247/MICE0247.pdf
http://hep04.phys.iit.edu/cooldemo/micenotes/public/pdf/MICE0253/MICE0253.pdf
M.Ellis - CM24 - 3rd June 2009
4
We Need to Turn this:
Conditions
Database
(EPICS)
Controls &
Monitoring
network
RAW
MICE
DAQ
Transfer
Box
Online
Buffer
(MICEACQ05)
2 TB / day
= 200 Mbps
RAW
Online
Farm
Online
Reconstruction
MiceNet
DAQ network
CASTOR
tape
Offline
Reconstruction
Tier 1
network
CASTOR
disk
UK GridPP
Tier 2
Farms
Analysis
“Chaotic” (on-demand) analysis
Generic user certificate
LFC
Any
Tier 2
Farm
MC
simulation
Tier 2 SE
disk
results
M.Ellis - CM24 - 3rd June 2009
Figure 1: Data flow from the MICE experiment.
Short-dashed entities require confirmation.
Long-dashed lines represent borders between subnets.
User local
disk
5
Into This:
Conditions
Database
(EPICS)
Backup link
to Tier 1?
Failover link
to Tier 2 ?
Controls &
Monitoring
network
VOMS,
DNS,
NTP ?
ISIS or PPD
network
Create file
metadata?
RAW
MICE
DAQ
RAW
Online
Buffer
Transfer
Box
(MICEACQ05)
RAW
2 TB / day
= 200 Mbps
RAW
Online
Farm
Online
Reconstruction
Ephemeral
ROOT
histograms
Optical Fibre
1 Gbps to Tier 1
Configuration
Database?
“Robot”? certificate
VOMS “archiver”? role
MICE_RAW_TAPE? token
MiceNet
DAQ network
Config DB
“API”?
Offline
Reconstruction
?
?
RAW
MyProxy
CASTOR
tape
Tier 1
network
AMGA
Semi-automated process
UK GridPP
Tier 2
Farms
“Anointed user” certificate
VOMS “production” role
MICE_RECO_DISK? token
Analysis
“Chaotic” (on-demand) analysis
Generic user certificate
CASTOR
disk
RECO
(ROOT trees)
LFC
BDII
RECO
?
Any
Tier 2
Farm
MC
simulation
Tier 2 SE
disk
results
archive?
results
M.Ellis - CM24 - 3rd June 2009
Figure 1: Data flow from the MICE experiment.
Short-dashed entities require confirmation.
Long-dashed lines represent borders between subnets.
User local
disk
6
“Known Unknowns”
 Meta-data:


First list of potential search terms.
What technology/DB to use?
 LFC namespace:


Proposed at last CM.
Need to agree (or disagree!) to it now so that we can start
registering data to the GRID.
 MICE Data:



Volume
Lifetime
Access control
 Proposed to arrange a 1 day workshop for
interested / key people to settle these points.
M.Ellis - CM24 - 3rd June 2009
7
G4MICE Status & Plans
 G4MICE 2.0.0 was released on the weekend.
 It is the first release with the new compiler and
external libraries:







gcc 4.3.2
GEANT4 9.1.p01
CLHEP 2.0.4.2
ROOT 5.22.0
GSL 1.12
Unpacking 2.0
RecPack 20090424
 It compiles without warnings on 32 bit and 64 bit
machines and this will be enforced for all future
releases.
M.Ellis - CM24 - 3rd June 2009
8
Tests
 We currently have:
 Unit Tests: 81
 System Tests: 55
 Integration Tests: 0 (several written, but none yet added
to testing script).
 In total, the testing is currently checking over 13,000
simulated, digitised, reconstructed values or other output
from various parts of G4MICE.
 Still need Unit Tests for the RealData domain. Will work on
these over the summer.
 Need to add System/Integration tests for the common use
cases.
 Please let me know how YOU use G4MICE so I can add some
tests to ensure that what you do continues to work the way
you need it to.
M.Ellis - CM24 - 3rd June 2009
9
Detector Reconstruction
 Tracker reconstruction has been improved as a
result of Aron’s analysis of the KEK data and the
ongoing cosmic ray tests in lab 7.
 Two areas to be improved before the next
production release:


Remove hard coded tracker number in SciFiDigit class.
Fix automatic choice of PR algorithm based on the magnetic
field in each tracker.
 TOF reconstruction is well advanced (see reports on
work by Mark, Yordan and Vassil).
 CKOV reconstruction is also coming along nicely
(David Adey and Vassil).
M.Ellis - CM24 - 3rd June 2009
10
Next Steps
 Add tests for the FileSplitter and FileMerger
Applications.
 Ensure all Recon Unit tests are working correctly
and increase the scope to cover cosmic ray as well
as MICE physics cases.
 Generate release with all of the above for
installation on the GRID.
 In parallel with the above:


Produce a 64-bit version of the compiler and external
libraries for installation on 64 bit machines.
Produce a Mac version of G4MICE (growing demand).
M.Ellis - CM24 - 3rd June 2009
11
Concerns
 As mentioned at a past VC, I am still concerned
about the lack of direction given to those working
on the code for the PID detectors.
 Both Mark and Yordan have been working on the
TOF, and I fear this has resulted in
wasted/duplicated effort.
 There would appear to be a similar potential in the
CKOV with Vassil and David.
 I would hate to think that important tasks (e.g.
implementing the actual MICE hall geometry in the
software) is suffering due to lack of mouse-power
while effective mice are duplicating the work of
others.
M.Ellis - CM24 - 3rd June 2009
12
Activities that need an Owner
 There are a number of key areas for which
someone needs to take “ownership”:



MICE hall geometry (I am starting on this with Ash
Jamdagni and the help of people at RAL, but this will not
move very quickly).
Production – this was discussed during Henry’s talk
yesterday, someone (or a group) needs to take charge of
the eventual offline reconstruction of our data (we already
have lots to practice on!).
PID related software – unlike the tracker group, there still
does not appear to be any clear leadership for the other
detectors (with the possible exception of the EMR).
M.Ellis - CM24 - 3rd June 2009
13
Any Volunteers?
M.Ellis - CM24 - 3rd June 2009
14