TDS/PDS Discussion 14 Aug 2002

TDS/PDS Discussion

Problems

What should the content of the PDS be, given its (presumably) dual role as intermediary for the TDS, and as a source of outside-of-Gaudi analysis?

We need to regenerate the "analysis" ntuple - akin to the ntuple used for PDR analysis last year.

Consequences of TDS/PDS

PDS must contain what the TDS needs to be reconstituted. For digi and recon input, at least.

This is included in what we have implemented, but the PDS is a superset of what is needed to carry on the next step. I think this is largely true for the TKR. Recon output is not restricted by this requirement yet, though it might have some mild ones when we add the Event interpretation and ID step.

Analysis

Analysis is possible in both userAlg - with direct access to the TDS - and via "command line" Root (or any other tool capable of reading Root files).

userAlg analysis

(currently) uses TDS for input, with output via AIDA histograms (converted to Root by Gaudi) and user ntuples.
potentially has access to Gaudi services, like detModel and G4/RcParticle
intrinsically better access to data/algorithms
suffers the CLG cycle
could be augmented to allow access to Root trees (but in some sense, what is the point - if there is no use of the TDS and Gaudi utilities?)

Root analysis

no access to Gaudi services (or it is difficult at present to supply it). What you see in the Root file is what you get.
lower turnaround time for analysis than in Gaudi. For rapid analysis prototyping.
is there an issue about available algorithms to manipulate the data in Root?

Further details

Full recon ought to have event interpretation and "ID", including background rejection - we will need this to create the summary L1 database in the fullness of time. Perhaps this is merit?

Bill Atwood suggests that Recon "analysis" output

not require any significant geometry to interpret
be able to (re)do event interpretation and ID.
he has suggested a number of quantities to go into the output that are not needed by the TDS.

Does this lead to a mix of PDS/TDS + analysis quantities?

Action: Analyze the current PDS + Bill's suggestions to see if they can be classified as TDS or analysis entities.

Ntuples

We will also need an afterburner to extract the quantities needed to create the "analysis" summary ntuple. It probably should not have to do any real thinking - just minor rearrangements of code on the TDS. And can be performed from a single algorithm, with access to inputs via the TDS.

Action: Post today's AnaGroup meeting, Steve will poll the analysts for their desires from the ntuple, to come back next week with a first cut.

R. Dubois Last Modified: 01-Jun-2010 15:47:40 -0700