Upgrading The Huawei Cloud Storage Benchmark Framework For Root6 Compatibility
Surya Seetharaman, Maria Arsuaga Rios, Seppo Heikkilä · INFM-OAR (INFN Catania) · 2015
Project Specification and Motivation The multi-client benchmark framework of the Huawei Cloud Storage system at CERN was using ROOT5 to provide an analysis about the collection of data retrieved by the clients. Recently a new version of ROOT (ROOT6) was released with many improvements. Hence, to use the benchmark framework more effectively, it had to be upgraded so that it could run using ROOT6. Also this benchmark framework is generic in the sense that this upgraded benchmark framework can be used for measuring the scalability of other storage technologies at CERN like Kinetic Drives or EOS. As a part of the Data and Storage Services group of the IT department (IT-DSS) at CERN, the task was to upgrade the cloud storage benchmark framework and run the existing tests on the upgraded benchmark. Abstract The European Organisation for Nuclear Research (CERN) is the largest research center in the world in the field of particle physics. Huge amounts of data, in the order of Peta Bytes, are being generated from the High Energy Physics experiments going on at CERN. The massive data growth prompts CERN to evaluate new storage technologies, to store and analyse the data that is being generated from the experiments. The scalability of the storage system is important for CERN as the laboratory faces the ever increasing demands of its physics users. Huawei, a CERN collaboration member, focuses on investigating the applicability of new storage techniques and architectures to the storing of high energy physics data from the LHC eperiets. I earl , Huaeis loud storage sste as deliered to the CERN site, and in three months, the installation and benchmark performance evaluation ere opleted. Huaeis loud storage proed to sho horizontal scalability in writing and reading performance and capability to automatically recover from storage node failures. One of the main objectives of the Huawei - CERN partnership is to investigate the performance and scalability of a new cloud storage system and compare the results with more traditional approaches used at CERN. To perform this quantitative study, a multi- client benchmark has been developed using ROOT to allow statistical data analysis. Currently, a new version of ROOT has been released with many improvements. In order to allow for more efficient testing and rapid deployment of new benchmark workloads, the existing multi-client benchmark had to be upgraded to be compatible with ROOT 6. This report documents and justifies that the benchmark framework was upgraded and verified to be able to produce the same results as the previous benchmark framework before the upgrade.