Hello Hassen,
I think a major obstacle in using HDFS and Hadoop with ROOT files is that HDFS is designed for large sequential reads and AFAIK does not allow seek()ing. I think (at least by default) ROOT relies on being able to seek() to read TFiles and also TTrees. You can change how baskets are organised in a TTree so that you get all branches of event1, event2, event3, ..., eventN stored in the file instead of storing branch after branch in the file. Not sure if this would eliminate the need to seek().
My two cents,
Tim
-- http://tim.jottit.com/Received on Thu Apr 19 2012 - 13:40:03 CEST
This archive was generated by hypermail 2.2.0 : Thu Apr 19 2012 - 17:50:01 CEST