J-PARC (Japan Proton Accelerator Research Complex) consists of much equipment. In the Linac and the 3 GeV rapid cycling synchrotron ring (RCS) in J-PARC, data of about 64,000 EPICS records have been collected for control of these equipment. The data volume is about 2 TB every year, and the total data volume stored has reached about 10 TB. The data have been being stored by a Relational Database (RDB) system using PostgreSQL since 2006 in PostgreSQL, but it is becoming that PostgreSQL is not enough in availability, performance, and flexibility for our increasing data volume. We are planning to replace PostgreSQL with Apache Hadoop and Apache HBase to accumulate enormous operation data produced from the Linac and the RCS in JPARC. HBase is so-call NoSQL, which has scalability to data size at the cost of the high broad utility of SQL. HBase is constructed on a distributed file system provided by Hadoop, a cluster with advantages including automatically covering its cluster nodes' breakdowns and easily adding new nodes to expand its capacity. The new database system satisfies high availability, high performance, and high flexibility of storage expansion.
展开▼
