Brahms Participation in Mock Data Challenge

Brahms Participation in Mock Data Challenge (MDC)

This page is a copy of the 2-3 slides I showed on February 5 at the RCF local meeting in regard to the BRAHMS plans for the MDC. It has that BRAHMS will adapt ROOT officially as will likely happed soon. For more information go back at look at the other references.

Goals of Mock Data Challenge.

Implementation and operation of analysis chain from generation of simulated events, digitization to track reconstruction, Particle Identification and simple physics analysis.
Test the Full chain of software. Learn and understand how to operate this in an environment of multiple (jobs) users, and different access to data.
Employ software that will (or very likely to) be used in the production analysis code for Brahms.

MDC1 Specifics

Detectors included TPCs, DC, TOF, (Beam-Beam)
Tracking with existing code
PID for TOF
Event selection with beam-beam and PID start.
Invariant spectra (pions, protons)

This set of detectors selected due to

existence of code
Importance for analysis chain.

Tasks to be done in preparation and during the MDC.

To get meaningful statistics for single angle, magnetic settings one needs in the order of 1000 tracks. This can for non-central and not so forward angles be a considerable amount of events which has to be tracked. The time needed for this depends on the completeness of the Geant simulation.

For the main purpose of the software and stress testing one can turn several elements off in the simulations to speed up this part. Otherwise with events in full phase space and with both spectrometers & DX included the time per event on the rpro’s is in the order of 5 minutes per event. As a first rough estimated the total number of events needed is

4 angle settings * 3 field settings * 1000 events / <0.1 tracks/event> = 120,000 events.

This is roughly 2.5 Gb worth of data (and approximately 20 min worth of real data).

It would take ~10,000 hours to generate this with full phase-space events in Geant. This might be possible, and we will attempt a cheaper solution as outlined above. Still it is likely to require a couple of CPU months for this task.

Event generation - This is a fairly trivial task; for the small solid-angle detectors the same simulated events can be used multiple times by coordinate transformation to effectively give a large event sample. Needs 2-5K of central and Min bias Au+Au data. (1-1.5 Gb of data)

Digitized data - This requires running GBRAHMS multiple times for several angle, and magnetic field settings.

CPU requirements to generate?
Disk requirements [120,000*30kB 2-3 Gb]

Tracking TPCs and DC - This is the most cpu intensive part

Disk output needs 2/3 of input (2 Gb)
CPU requirements [400 kSpecInt95.sec]

PID, Beam-Beam Counters - This has still to be fully written. The cpu needs are modest, but i/o rates reasonable high.
Cross section Code - same status and consideration as above.

Time scale for implementation

Tracking and PID implemented by April 98
Cross section code and global detector analysis ready by July 98.

MDC1

Framework will use ROOT classes. The i/o will be flat files, possibly root files though not determined. DST will be ROOT files

The overall file sizes need are small so in order stress test the RCF system it is therefor necessary to run the same multiple input files (or duplicates thereof).

Calibration access will most likely by files, and not by direct access to Database (Obj-y).