Simulation-Based Performance Analysis of the ALICE Mass Storage System

CERN – the European Organization for Nuclear Research today, in the era of big data, is one of the biggest data generators in the world. Especially interesting is transient data storage system in the ALICE experiment. With the goal to optimize its performance this paper discusses a dynamic, discrete event simulation model of disk based Storage Area Network (SAN) and its usage for the performance analyses. Storage system model is based on modular, bottom up approach and the differences between measured and simulated values vary between 1.5 % and 4 % depending on the simulated component. Once finished, simulation model was used for detailed performance analyses. Among other findings it showed that system performances can be seriously affected if the array stripe size is larger than the size of cache on individual disks in the array, which so far has been completely ignored in the literature