Development and tests of the event filter for the ATLAS experiment

The trigger and data acquisition (TDAQ) system of the ATLAS experiment comprises three stages of event selection. The event filter (EF) is the third level trigger and is software implemented. Its primary goal is the final selection of interesting events with reduction of the event rate down to ~200 Hz acceptable by the mass storage. The EF system is implemented as a set of independent commodity components sub-farms, each connected to the event builder subsystem to receive full events and on the other side to the sub-farm output nodes, where the selected events are forwarded to mass storage. A distinctive feature of the event filter is its ability to use the full event data for selection directly based on the offline reconstruction and analysis algorithms. Besides the main duties on event triggering and data transportation, the EF is also able to provide additional functionalities, like monitoring of the selected events and online calibration of the ATLAS detectors. Significant design improvements are currently under development to provide these additional functionalities to the EF system. The event filter was deployed and tested in data triggering at the ATLAS combined test beam at CERN in 2004. The EF is also subject to various tests on dedicated test-bed computer clusters, where both the EF functionalities and performance are studied. Special tests are being done when running the EF software components on the large-scale EF farms, with hundreds of computers. This paper describes the current activities and plans on development and testing of the ATLAS event filter system