关键词: 云存储,一致性哈希,层次分析法,副本分布,多目标优化

Abstract: The data distribution strategy in cloud storage systems has shortcomings that the distribution strategy only considers the load balancing among physical storage nodes instead of the heat load balancing,waiting time and other factors.Aiming at this question,a rational and efficient data distribution strategy local optimum distribution (LODS) was proposed.By defining a series of new definitions to describe the correlations in LODS,employing the hashing method to act as the core ideal and presenting a couple of multi-objective optimization rules,LODS combines the above innovations and the AHP method to give the finally optimized storage scheme.The proposed storage scheme can search the target storage node in cloud system within a local range with a optimized decision radius.The decision radius is stable without variation while the large-scale storage amount is tested.Once the decision radius is “0”,what the LODS degenerates is almost corresponding to that of Amazon S3.Experiments show that the average advantage of the proposed cloud storage strategy overs that used in HDFS,Amazon S3 in terms of the strategies of storage load balancing,load balancing heat and waiting time.

Key words: Cloud storage,Consistent hashing,AHP,Data distribution,Multi-objective optimization