Performance of the ATLAS DAQ DataFlow system

The baseline DAQ architecture of the ATLAS Experiment at LHC is introduced and its present implementation and the performance of the DAQ components as measured in a laboratory environment are summarized. It will be shown that the discrete event simulation model of the DAQ system, tuned using these measurements, does predict the behaviour of the prototype configurations well, after which, predictions for the final ATLAS system are presented. With the currently available hardware and software, a system using ~140 ROSs with 3GHz single cpu, ~100 SFIs with dual 2.4 GHz cpu and ~500 L2PUs with dual 3.06 GHz cpu can achieve the dataflow for 100 kHz Level 1 rate, with 97% reduction at Level 2 and 3 kHz event building rate. ATLAS DATAFLOW SYSTEM The 40 MHz collision rate at the LHC produces about 25 interactions per bunch crossing, resulting in terabytes of data per second, which has to be handled by the detector electronics and the trigger and DAQ system [1]. A Level1 (L1) trigger system based on custom electronics will reduce the event rate to 75 kHz (upgradeable to 100 kHz – this paper uses the more demanding 100 kHz). The ________________________________________ #. Also affiliated with University of California at Irvine, Irvine, USA *. On leave from Henryk Niewodniczanski Institute of Nucl. Physics, Cracow +. Presently at CERN, Geneva, Switzerland 91 DAQ system is responsible for: the readout of the detector specific electronics via 1630 point to point read-out links (ROL) hosted by Readout Subsystems (ROS), the collection and provision of “Region of Interest data” (ROI) to the Level2 (L2) trigger, the building of events accepted by the L2 trigger and their subsequent input to the Event Filter (EF) system where they are subject to further selection criteria. The DAQ also provides the functionality for the configuration, control, information exchange and monitoring of the whole ATLAS detector readout [2]. The applications in the DAQ software dealing with the flow of event and monitoring data as well as the trigger information are called “DataFlow” applications. The DataFlow applications up to the EF input and their interactions are shown in Figure 1. Figure 1 ATLAS DAQ-DataFlow applications and their interactions (up to the EventFilter) SFI L2PU L2SV DFM pROS ROS ROI data