Evolution of the ReadOut System of the ATLAS experiment

The ReadOut System (ROS) is a central and essential part of the ATLAS data-acquisition system. It receives and buffers event data accepted from all sub-detectors and first-level trigger subsystems. Event data are subsequently forwarded to the High-Level Trigger system and Event Builder via a GbE-based network. The ATLAS ROS will be completely renewed in view of the demanding conditions expected during LHC Run 2 and Run 3. The new ROS will consist of roughly 100 Linux-based 2U-high rack-mounted server PCs, each equipped with 2 PCIe I/O cards and four 10GbE interfaces. The FPGA-based PCIe I/O cards, developed by the ALICE collaboration, will be configured with ATLAS-specific firmware, called RobinNP. They will provide connectivity to about 2000 point-to-point optical links conveying the ATLAS event data. This dense configuration provides an excellent test bench for studying I/O efficiency and challenges in current COTS PC architectures with non-uniform memory and I/O access paths. In this paper the requirements for Run 2 and the design choices for a system complying with or exceeding them are described. The results of performance measurements for different computer architectures, highlighting the effects of non-uniform resource distributions, are discussed. Finally the status of the project and outlook for operation in 2015 are presented.