PAT Workflow

The PAT workflow is divided in a series of layers as shown on the figure below:

• Layer 0 performs the "cleaning" tasks on the standard reconstructed object (in AOD or RECO formats), computes related information (e.g., Monte Carlo and trigger matching), and takes care of the low-level handling of references.
• Layer 1 collapses the output of layer 0 into compact objects containing all the necessary information.
• Layer 2 performs event-hypothesis dependent tasks: this is where the analysis starts.

The PAT workflow

The PAT consists of framework plugin modules: cleaners for layer 0, producers for layer 1. These plugins themselves call standalone algorithms and helper classes.

Notes:

• all classes are defined in the namespace pat::;
• the configuration files can be found in CMS.PhysicsTools/PatAlgos/python, ordered by task.

Tasks performed inside the PAT

The tasks performed insode PAT are:

• Monte Carlo and trigger matching
• Lepton isolation and particle identification
• Objects cleaning (e.g., jet-electron overlap removal)
• The definition of global event variables (e.g., total event charge, object multiplicity)

They are represented in the following structure:

• Layer 0: cleaning and association on AOD content
  • More details about the PAT cleaning and the Monte Carlo and trigger matching
• Layer 1: production of PAT objects from information gathered at Layer 0.
  • More details about PAT data formats and PAT trigger information
• Layer 2: easy access to event hypothesis structures using Layer 1 objects.
  • More on event hypothesis

-- RogerWolf - 13 May 2009