Parallel database systems and its algorithms

Parallel database systems are becoming increasingly popular because of the throughput and scalability it provides. This paper describes the various ways to achieve parallelism in database. There are different parallel database architectures and shared memory and shared nothing architecture are most extensively employed in various applications. The execution of query can be controlled in either control flow way or data flow way. This paper describes the control flow approach and data flow approach. Large number operations that are performed in parallel database are sorting and joins. We illustrate various sorting and join algorithms and focus on hash based join algorithms which are most efficient. This paper also addresses the major issues such as overloaded processor and communication overhead. Finally we describe the declustering mechanism and performance tuning used to overcome the above issues.