Performance and evolution of the ATLAS TDAQ system with pp collisions at 7~TeV

During the data-taking period from 2009 until 2011, the ATLAS TDAQ system has been used very successfully to collect proton-proton data at LHC centre-of-mass energies between 900 GeV and 7 TeV. The TDAQ system is mostly made of off-the-shelf processing units organized in a farm of 2000 elements. The trigger system is designed in three levels reducing the event rate from the design bunch-crossing rate of 40 MHz to an average recording rate of about 300 Hz. Using custom electronics with input from the calorimeter and muon detectors, the first level rejects most background collisions in less than 2.5 $mu$s. The two following levels are software-based triggers with average decision times of 40 ms and 4 s respectively. The trigger system is designed to select events by identifying muons, electrons, photons, taus, jets, and B hadron candidates, as well as using global event signatures, such as missing transverse energy. In 2011, the TDAQ system has been operated with an overall efficiency of 94%, while meeting evolving and demanding conditions. With the LHC peak luminosity increase through 2011, the scalability and operational margins in terms of bandwidth and dataflow have been stressed. During the heavy ion runs in 2011 the system was operated at the limit of the installed computing power, enabling the evaluation of the effectiveness of the current installation and the validation of the operation modeling tools. We give a description of the system together with the operational experience with an emphasis to the data-taking in 2011. We also give an overview of the performance of the different trigger selections. Distributions of selection variables used by the different trigger selection are shown and compared with the offline reconstruction. Examples of trigger efficiencies with respect to offline reconstructed signals are presented and compared to simulation. These results illustrate a very good level of understanding of both the detector and trigger performance. Furthermore, we describe how the trigger selections have evolved with increasing LHC luminosity to cope with the increasing pile-up conditions. In addition, driven by the lessons learned from operation, the ATLAS TDAQ system current limitations together with the present strategies being put in place to solve them will also be described.