JungHyun Kim

28 October 2013
IEEE 2013 NSS

Abstract: We introduce the distributed computing and data handling systems at Belle II, a next-generation B factory experiment at KEK. The systems have to handle an amount of data eventually corresponding to 50 times the level of its predecessor experiment, Belle, by the end of 2022. The data size is comparable to or more than the ones of LHC experiments. For the Belle II experiment we employed the distributed computing system based on DIRAC. DIRAC provides many features that we would have to develop and an interface to grid and cloud resources. Furthermore, we need a reliable and efficient metadata service that translates selection queries into a set of files. We design our metadata service based on AMGA (Arda Metadata catalog for Grid Application). First, this year, the Belle II computing group realized a test bed based on DIRAC and has entered a new phase of testing the feasibility of our system. Second, in order to identify the right files that should be used as input for a project, we set up an AMGA metadata server which allows to get a list of files, in form of GUIDs. The metadata is extracted from the data files that are written to storage resources and used to fill the AMGA metadata catalog. Using the catalog a user query is translated into a file list and a list of sites where the data is available. This information is used to submit jobs that process the requested data. In conclusion, we have designed and constructed a Belle II distributed computing system and data handling system. It is essential for us to do practical tests for the operation of DIRAC, scalability of metadata service, scheme of job submission, and other involved components. The status of the Belle II computing system and Monte-Carlo production campaign are presented.