Features

A large number of functionalities have been implemented in PAF while trying to keep the configuration task as simple and automatic as possible. It has been developed with the idea of getting the best performance with an easy way to interact with it.

  • Modular analysis framework: In our latest release, we have introduced the concept of a project made of several analyzers providing easy communication abilities to all of the elements. This way specialized analyzers (ex: Electron selector, MET estimator,…) can be easily shared among different analysis developers.
  • PAF EnvironmentsProcessing environments: Including sequential processing, PROOF Lite and other dynamic PROOF cluster builders (PoD, PROOF Cluster, PROOF Cloud). To change from one to another just a single line needs to be edited. New environments can be easily added.
  • Transparent lazy loading of branches: A smart dynamic mechanism allows PAF to instruct ROOT to only load those branches used in the analysis resulting in processing rates up to 10 times faster depending on the complexity of the analysis.
  • Support for local or distributed package compilation: Through a single switch PAF can be instructed to compile analysis packages locally (ideal for homogenous environments) or in the slaves (needed in the case of heterogeneous clusters).
  • Logger: We have developed an extensible logger to improve PAF runtime information.
  • Dynamic histograms: One or several histograms produced in the analysis can be selected to be plot during the data processing.
  • Homogenous and efficient variable passing: A simple mechanism has been implemented to pass information from the main process to the analyzers, and  between analyzers.
  • Tree inspector: An independent tool to easily inspect and find out the content (branches) in a ROOT file The tool provides cut and copy code very useful in the development of a new analysis class.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.