Replicated check pointing is widely used to tolerate multiple node failure. Multiple node failure capability is enhanced by generating the replicas on demand. Replica on demand for mission critical data and their placement is a crucial task. Placement of such huge number of replicas at run time bottle necks the performance and dependability. A high overhead is due to recomputaion of placement from initial stage. The performance becomes more severe if large number of nods and their data need to replicate. In order to improve the dependability these replicas need to placed on totally distant set of nodes. This paper presents dynamic replication management architecture with improved performance and dependability. Improved performance is achieved by avoiding the recompuation for all replicas from initial stage and placement of replicas uniformly over all distant nodes. Improved dependability is achieved by placing the replicas over totally distant set of computing nodes. This paper proposed architecture as well as experimental result to show effectiveness of proposed approach.
展开▼