A persistent back-end for the ATLAS TDAQ online information service (P-BEAST)

This paper describes P-BEAST, a highly scalable, highly available and durable system for archiving monitoring information of the trigger and data acquisition (TDAQ) system of the ATLAS experiment at CERN. Currently this consists of 20,000 applications running on 2,400 interconnected computers but it is foreseen to grow further in the near future. P-BEAST stores considerable amounts of monitoring information which would otherwise be lost. Making this data accessible, facilitates long term analysis and faster debugging. The novelty of this research consists of using a modern key-value storage technology (Cassandra) to satisfy the massive time series data rates, flexibility and scalability requirements entailed by the project. The loose schema allows the stored data to evolve seamlessly with the information flowing within the Information Service. An architectural overview of P-BEAST is presented alongside a discussion about the technologies considered as candidates for storing the data. The arguments which ultimately lead to choosing Cassandra are explained. Measurements taken during operation in production environment illustrate the data volume absorbed by the system and techniques for reducing the required Cassandra storage space overhead.