Development of Simple Benchmark Tool for Data Acquisition Network

In data acquisition (DAQ) systems used for highenergy experiments in physics such as the ATLAS experiment at CERN, parallel processing is employed for efficient acquisition and analysis of experimental data from multiple source nodes. The architecture used to collect the event fragments is designed for many-to-one simultaneous communications for each event. The architecture is very simple but causes packet losses in the system without any flow controls, and with less fairness and less throughput of each flow. Many studies have tried to solve the problem by applying architectural and technological approaches. General-purpose network performance tools such as iPerf and Netperf were used to evaluate such DAQ networks. However, the traffic of a DAQ system is not general, because many source nodes transmit data to one destination simultaneously having the same timing. The tools are not able to evaluate the specific traffic correctly. We, therefore, developed a simple benchmark tool, DAQPerf, to evaluate the network for specific traffic easily. We measured the data transmission rates for the specified trigger rate in order to evaluate the performance of the DAQPerf, and concluded the DAQPerf could be used until 40 kHz of the trigger. And we also measured errors of data arrival rates for the specified trigger rates. As a result, we confirmed that the errors of them were from 8 to 19 sec. We confirmed DAQPerf was useful to evaluate the simultaneous many-to-one communication system such as a DAQ system.