Some Thoughts on
Goals of Working Groups
D. Olson
30 June 1997
Reminder of Project Goals
- Produce 50 TB dataset of typical experiment. Should mock
up 2-year running for STAR and PHENIX.
- Dataset should have "signals" which can be
queried at least at every order of magnitude in density.
Something every event (like STAR E-by-E characteristics),
something every 10 events like cascade or D or low-Pt
bubbles, something every 10**2, 10**3, 10**4, 10**5,
10**6, 10**7 events.
- Dataset should consist of DST, uDST and event-generator
(final physics event) type data.
- Ability to analyze new queries.
- Ability to re-structure (reorganize) 50 TB in xx? hours.
- Ability to query 50 TB dataset (on tape) and select (move
through analysis CPU's) 1 TB in 5 hours.
General Goals
- Define tests to do over the next several months in order
to make the choices for the final implementation.
- Identify tasks & assignments.
- Identify specific requirements.
- Assign responsibility to write up sections for the
project description document.
- Estimate manpower.
WG on Clustering and Object Store
- Consider implications with and without Objectivity /
RD45.
- Consider relationship between storage objects
(containers, clusters, files, tapes, ...) and physics
objects (events, tracks, points, hits, vertices,
particles, ...). How does this relationship get expressed
in C++?
WG on Simulations and Data Generation
- Event format for generators.
- Matrix of number of events to produce vs. source
(generator, geant, ...)
- Data volume and CPU time requirements.
- Determine amount of data to produce directly.
- Determine amount of data that can be
"duplicated" by randomizing the original in
some fashion.
- Determine signals to put into data.
WG on Object Model & Performance Monitoring
- What parameters of storage objects and physics objects
need to be monitored?
- What parameters can be monitored (via HPSS, via
Objectivity, via alternative)?
- What integration is required for prototype implementation
and final implementation?
WG on Proto-type Implementation & Facilities
- Describe I/O in STAF.
- Identify prototype scope (see note from
Soren).
- Identify required resources & facilities.
- Consider architecture of final implementation.
- Identify what is known about NERSC & RCF facilities.
- Consider SciTL
- Scientific Template Library.