Abstract:
The paper is dedicated to a problem of effective query processing in cluster database systems. An original approach to data allocation and replication at nodes of a cluster system is presented. On the basis of this approach the load balancing method is developed. Also, we propose a new method for parallel query processing on the cluster systems. All described methods have been implemented in "Omega" parallel database management system prototype. Our experiments show that "Omega" system demonstrates nearly linear scalability even in presence of data skew. © 2010 Springer-Verlag.