GRFM: an efficient grid-based replication and fault tolerant middleware

摘要

Grid computing provides geographically distributed resources for large-scale complex applications which involve huge data management services. There is a need for availability, fault tolerance, scalability and high-performance in a service-oriented environment as grid. The proposed grid-based replication and fault tolerant middleware (GRFM) is a solution for data availability by eliminating faults in grids. GRFM is an adaptive replication grid middleware framework especially designed for high-performance applications. It enables data synchronisation between multiple heterogeneous databases located remotely in a grid. The proposed framework offers: an effective replicating strategy, mechanism to handle faulty nodes, totally ordered communication mechanism, improved membership service, hash-based replica locating service, and support for scalability and security features. GRFM reduces the system overhead by reducing the traffic in the grid and thereby improving the performance of data management for large-scale complex grid applications. Experiments are carried out on well-known grid computing tool kit Aneka (Chu et al., 2007). The experimental results show that GRFM performs better when compared to universal description, discovery and integration framework (UDDI) lazy replication (Sun et al., 2004), WS-based ring replication protocol (WS-RRP) (Luckow and Schnor, 2008), and JGroups (Ban, 1998) total ordering and group communication service approaches on real-time transaction databases.