期刊名称：International Journal on Computer Science and Engineering
印刷版ISSN：2229-5631
电子版ISSN：0975-3397
出版年度：2012
卷号：4
期号：06
页码：1146-1152
出版社：Engg Journals Publications
摘要：The performance of checkpointing replication fault tolerance technique is severely bottlenecks due to handling of number of replicas generated for a large number of nodes to tolerate multiple faults such as multiple failure of nodes, processes etc. In fusion based approach, these checkpointing replicas stored at large number of computing nodes is aggregated into some data structure to handle efficiently through fused data structure. These impose higher overheads of fusing a large numbers of checkpointing replicas. In this paper, a self fused checkpointing replication (SFCR) for cloud computing is proposed. All checkpointing replicas assigned to store at a particular node are stored in a self-fused-shared-checkpointing-replicas file already created and located at every node rather than storing as a separate checkpointing element and than fusing. Thus, it eliminates the need of further fusing of the checkpointing relicas stored at different checkpointing replicas storage nodes, as checkpointing replicas assign to store a particular node stored in an already created fuse file at every checkpointing replicas storage node. It improves the performance without affecting the specified fault tolerant capabilities as failure of any node will result in loss of all replicas irrespective of separate checkpointing file or shared checkpointing fused file. Costs to maintain these set of self fused shared files are obviously less than the number of separate replicated files in terms of time and efforts that it takes to create, and update a file. Thus, proposed approach is enhancement of performance without compromising the specified fault tolerance capability. At the same time when system seems to be prone to many number of faults, some specific self fused shared files consist of important and critical data that can be further replicated to enhance the fault tolerant capability of a group of important and critical nodes or processes at run time. Thus, it also provides adaptive dynamic fault tolerance capability in a simple manner. Effectiveness of proposed technique is also presented
