Title: MAP
1MAP
- Tracey Pratt
- Liverpool University
- Introduction/Overview of MAP
- How to submit jobs
- Present labeling conventions
- Filesizes/Job duration
- Diskspace
- Flexibility of MAP
- Outlook for MAP CDF
- Thanks
Dec 2003
2Multi-Array Processor
- 224 operational nodes 400MHz processors
- 30 to come
- At present successfully running jobs using
- cdfSim,
- ProductionExe,
- runMC,
- stnutple_prod.exe/stnuple.exe
- Present code version 4.9.1
3Overview of how MAP works
Job requested
Tar distributed to all nodes
Tar is sent to compass node
Tar working directory
Submit job via mapxqs
Produced files are returned at end of job
Job executed on all nodes
Output on hep52-57 (compass node)
Output copied to matrix
Data to be analysed
4How to submit jobs
- At present via me!
- 1) First test on cdfjif1 and see if your job
works - 2) Then tell me
- tcl for simulation (FakeEv/Pythia/Herwig)
- events
- executables to be used cdfSim/ ProductionExe
/ runMC/ stntuple_prod.exe - version to be generated (5.2.0 etc)
- return files requested
Status of jobs and where to find data at
http//hep.ph.liv.ac.uk/tracey/map.html
5MAP labeling convention
224 operational nodes arranged with 30
processors per rack labeling convention
map02011st map030131st map0601120th
map0901210st map02012nd map0302 32nd
map023030th map0930240th This
is to be reviewed, and files numbered 1, 2.etc.
To understand output numbers!
6Filesizes and Job duration
40min1 hour overhead staggered start on 120
nodes 1 hour overhead staggered
start on 224 nodes
So per day (24 hours) could make 1840 Zee
events ? 300 MB
7Disk Space
Present disk space 3 x 500 MB 1.5 TB 1 x 1.8
TB 1.8 TB 3.3 TB To add 2 x 1.8
TB 3.6 TB (hep58 and 59)
Issues storage space now if empty 51-56 500
MB 3TB (53,55,56 not available) 57-59 1.8 TB 3
5.4 TB (5859 not on) In use 1.5 TB 1.8 TB ,
1.8 TB2 to come soon! Capability per year
8Flexibility with MAP
Width
200
Mass (GeV)
300
400
generate different mass and width on each node
in one job submission And name files!
K/MPL 0.1
0.2
0.3
9Outlook
Continuous improvements being made e.g.
labelling from 1?224 with record of node add in
the extra nodes increase disk space
available update code versions available
Aim
Produce CDF MC data needed for Liverpool (CDF)
analyses
10to Tony!
Thanks
11The End!!