|
Short term plans for the GIOD Project (April
1998)
- Install on the CMS Work Group Server the prototype GIOD
analysis software which comprises:
CMSIM modules to
generate custom ASCII event files
C++ and Fortran application code to read event files,
populate Objectivity database with raw events, and
reconstruct raw event objects to track, cluster,
electron, photon etc. objects.
Java applet application code to read Objectivity
database and create event displays with tracks, clusters
etc..
Skeleton analysis framework software for the above,
including command driven module selection and ordering.
- Modify the Java event viewer applet by
- Rewriting using the Java 3D API, recently
released by Sun. Make the geometry, hits, tracks
and clusters as 3D objects that can be directly
manipulated in the display.
- Adding Java track fitting code that is a clone of
the C++/Fortran algorithm, and measuring the
performance of the JIT compiler when executing
this compute intensive task.
- Improve the track fitting algorithm used in the
C++/Fortran application.
- Carry out large scale simulations on the Exemplar for the
NPACI-approved Higgs --> gg studies.
- On the Exemplar, fully simulate >200k events for
studies of the QCD background to the Higgs.
- Liaise with CACR to have disk box connected to Exemplar.
- Make up new disk box for attachment to C200 ... and have
CACR NFS mount the filesystems on the Exemplar.
- Generate several thousand Higgs signal and background
events on the Exemplar (or C200) that can be used to
populate a large Objectivity database with raw and
reconstructed LHC data. These "realistic"
events to be used for performance studies of the
reconstruction/analysis framework, together with
feasibility studies for multiple client access across the
C200-Exemplar ATM
- Replicate the database containing these several thousand
events to the CMS Workgroup Server. Attempt "an
analysis" across the WAN, by running an analysis
task at Caltech that accesses the database at CERN, and
vice versa.
- Using ModNet, create regional centre models, insert the
parameters defined/guessed by Capiluppi, and run
simulations of the various CMS computing tasks. Attempt
to check the validity of the simulations by comparing
with simple "back of the envelope"
calculations. Identify bugs or problems in ModNet, and
suggest corrections or enhancements. Compare with the
software of Von Praun, and see how the two programs might
be combined. Estimate the effort, and wisdom, of
converting ModNet to Java ...
|