Massive Slow Down of LHCb upgrade HLT1 sequence running with TImeLineSvc on

Throughput tests for the LHCb upgrade relies on the TimeLine Service. We run Brunel multithreaded with Hive. Since commit bcfd63db in Gaudi, we started to observe that the endTime in the timeline service file was always set to 0 (Issue #14 (closed) ). The problem has been fixed in !610 (merged) , nevertheless this leads to wrong results. Also the throughput measured move from 6000 evt/s/node to 200.

In attachment the HLT1 sequence timing distribution among algorithms from Gaudi at bcfd63db commit and the one obtained with the current master.

Can we ping down what is going on? @ishapova , @sponce , @clemenci, @graven

Edited Feb 19, 2018 by Renato Quagliani

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information