Design for the distributed data locator service for multi-site data repositories

摘要

The Remote Experimentation Centre (REC) in Japan has been preparing to replicate the full dataset of ITER over 10 000 km distance. In such a multi-site data repository environment, the data location informing service will be essential to find and retrieve the data efficiently. Considering the long latency time and the self sustainability of remote sites, the data location database should be running at each repository site. Multi-master asynchronous replication between cooperating databases will be essential to realize the remote experimental collaborations in fusion research. This study has investigated the functional differences of some relational databases and found that Postgres BDR has the expected database replication capabilities. Bi-directional replication (BDR) tests by using the LHD database and SNET revealed that the throughputs are sufficient for remote collaborations in fusion experiments.