Speaker
Johannes Ebke
(Ludwig-Maximilians-Univ. Muenchen (DE))
Description
Historically, HEP event information for final analysis is stored in
Ntuples or ROOT Trees and processed using ROOT I/O, usually resulting in
a set of histograms or tables.
Here we present an alternative data processing framework, leveraging the
Protocol Buffer open-source library, developed and used by Google Inc.
for loosely coupled interprocess communication and serialization.
We save event information as a stream of Protocol Messages, which can be
read and written using high-performance code generated by the Protocol
Buffer software. No seeks are performed in write mode, and during
processing, making easy deployment over streaming network connections
possible.
The performance of our code on an example mock-physics analysis is then
compared with a ROOT analysis on the same data, showing the gain
obtained by leveraging current developments from outside HEP.
| Student? Enter 'yes'. See http://goo.gl/MVv53 | yes |
|---|
Primary author
Johannes Ebke
(Ludwig-Maximilians-Univ. Muenchen (DE))
Co-author
Peter Waller
(University of Liverpool)
