Trigger and Data Acquisition System
ATLAS is designed to observe up to one billion proton-proton collisions per second, with a combined data volume of more than 60 million megabytes per second. However, only a few of these events will contain interesting characteristics that might lead to new discoveries. To reduce the flow of data to manageable levels, ATLAS uses a specialised multi-level computing system - the Trigger System - which selects events with distinguishing characteristics that make them interesting for physics analyses.
The trigger system selects 100 interesting events per second out of 1000 million total. The data acquisition system channels the data from the detectors to storage.
The ATLAS trigger system carries out the selection process in three stages. The Level-1 trigger works on a subset of information from the calorimeter and muon detectors. The decision to keep the data from an event is made less than two microseconds after the event occurs, and the event is then retrieved from pipelined storage buffers. Of 40 million bunch crossings per second, less than 100,000 are kept by the Level-1 trigger.
The Level-2 trigger is a large array of custom processors that analyse in greater detail specific regions of interest identified by the Level-1 system for each event. In the mean time, the full event data is collected into buffers. A few thousand events per second pass Level-2, and have their data passed on to Level-3.
The Level-3 trigger is a large farm of CPUs which perform a detailed analysis of the full event data. About 200 events per second are left after the Level-3 analysis, and these are passed on to a data storage system for offline analysis.