Real-time parallel processing in high energy physics: architecture of the Blitzen data acquisition system

Data acquisition and processing systems for particle physics experiments in new high energy colliders (the Large Hadron Collider at CERN in Europe and the Superconducting Super Collider in the US) impose extreme demands on communication and processing systems. There is a requirement for resources to acquire interesting data from the vast quantity produced during experiments, to find events of interest in the data such as particle tracks and peaks of energy, and to analyze the physics of the experiments. Real-time response with 15 ns collision intervals is achieved through embedded, heterogeneous, hierarchical architectures with highly parallel processing components. The data acquisition computing environment is described. Emphasis is placed on the design of a prototype parallel processing system for the middle level of the hierarchy. Key features are the Blitzen parallel processing chip, HiPPI and VME interfaces, and logic to reformat data between heterogeneous devices.