Dynamic file migration to support parallel database systems

Research into DBMS (Database Management System) parallelism has been carried out to address the performance problems experienced in areas such as Decision Support. Distributed shared memory can alleviate the porting of commercial DBMSs to parallel platforms. However, this restricts the scaleability and performance achievable using shared nothing approaches. Consequently we propose the use of dynamic data migration techniques to support such parallel DBMSs. Using simulation techniques, we examine the effects which parameters such as migration threshold, migration block size and degree of data sharing have on file system performance. We present simulation results, providing useful insight into file system design.<
>