查看文章

The Belle experiment [1] started taking data in 1999 to confirm the Kobayashi-Maskawa theory [2] and accumulated an integrated luminosity of more than 1 ab− 1 during its 10-year operation. It was supported by a centralised computing facility at KEK, where almost all data processing, Monte-Carlo (MC) production and physics analysis was performed. The system was well designed for the amount of data and worked smoothly.

Now we plan the Belle II experiment, a next-generation B factory experiment at KEK, to search for new physics beyond the Standard Model which explains the behavior and interactions between elementary particles and aim to start in 2014. The Belle II computing system has to handle an amount of data eventually corresponding to 50 times the Belle level by the end of 2020. This means an amount of raw data of the order of 1010 events per year. In order to achieve the physics goals, it is required that the raw data is processed without any delay to the experiment data acquisition, in addition to the production of MC events corresponding to at least 3 times of beam data. Moreover, computing power for physics analysis has to be provided. The computing resources required for these purposes increase faster than the projected performance of CPUs and storage devices. Under this situation, it can not be expected that one site will be able to provide all computing resources for the whole Belle II collaboration. To solve this problem, a distributed computing model based on the grid is adopted. Figure 1 illustrates the concept of the Belle II computing model. Basically, KEK will host the main center that is responsible for raw data …