A survey of Shared-Nothing Parallel Database Management Systems

Distributed database systems can be implemented in a many different ways. Mostly, they are customized for a special environment to handle big data problems. The data warehouse sector relies on these amounts, but has changed from a data storage to a real time management support during the last years [3]. The resulting increase of compution and storage capacity poses new requirements to the database systems. Previous approaches of a parallel database environment tried to solve this problem with shared disk and memory approaches. The main contribution of this paper is the presentation of the current technology in the shared-nothing database sector. The concepts of the manufacturers Teradata, Greenplum and Netezza will be discussed for data warehouse requirements. Based on an architectural overview is a detailed insight of the index functionality given which is a crucial performance factor. Also data distribution algorithms of the manufacturers are analysed under data warehouse conditions. At the end is a comparison to other shared concepts (shareddisk, shared-everything) given and the question raised, if the actual approach can be fulfilled by the manufacturers.