Framework for data intercommunication and control of ATLAS High Level Trigger algorithms

The ATLAS experiment at the LHC is equipped with a sophisticated trigger system capable of reducing the 40 MHz LHC collision rate to the 70 kHz rate at the Level 1 hardware trigger and to the average 400 Hz rate at the High Level Trigger (HLT) during the Run 1 data taking in 2010-2012. In the HLT, the Steering Framework manages a few hundred of trigger algorithms: it evaluates every collision event and makes an accept/reject decision using as few resources as possible. Communication among algorithms is facilitated by a data navigation structure, implemented as a directed acyclic graph structure, that allows HLT algorithms fast discovery of detector regions containing interesting physics objects, as well as selecting an optimal way to build and traverse the event graph structure. For ATLAS running after the 2013-2014 shutdown (Run 2), more detailed trigger information is given to physics analysis users thus improving capabilities of trigger aware analyses. To reduce size of event data stored for physics analyses in Run 2, a new algorithm has been implemented that prunes the trigger navigation graph to an acceptable size, while retaining sufficient information for the physics analysis. This poster presents the design and operational experience of this framework for data intercommunication and control of HLT algorithms, and prospects for future development.