Research On Efficient Fault-tolerance Transition Mechanism For Multi-cloud Storage System

Cloud storage is widely used because of its convenient,flexible,cost-effective and extensible features.More and more enterprises and organizations transfer data from local to a single cloud storage service.But due to failure in the cloud,single cloud storage can not guarantee the requirements of high availability,while users need to migrate data,it will produce expensive costs.Therefore,the academic community proposed multi-cloud storage architecture,to alleviate those problems which are single cloud failure and cloud service provider lock-in,and quickly recognized and deployed by the industry community.On the other hand,the current distributed storage system store data with multiple replications to ensure the reliability and performance on read and write,after the data become cold,and store data with erasure coding to improve storage efficiency;The industry community calls this "multiple replications--erasure coding" mechanism transformation triggered by data temperature change as "Fault Tolerance Transition".At present,academia and industry community are still lack of the relevant research on the fault tolerance transition mechanism under the multi-cloud storage architecture.Therefore,this paper introduces the fault tolerance transition mechanism into the multi-cloud storage architecture,and constructs the corresponding prototype system ClouT(Cloud with Transition),stratifies the hot and cold data,uses the double replications to store the hot data,and uses the RAID5 for the cold data Storage,the fault tolerance mechanism of hot and cold data can tolerate a mistake,so as to ensure the reliability of the data after transition.When the data occurs the transition between hot and cold,and need to execute download,coding and other operations for data,which will produce a lot of bandwidth overheads and costs.In order to reduce the bandwidth overhead and cost in the transition process,we optimize the fault tolerance transition mechanism in ClouT.The intermediate node coding and the delay-based batch transition algorithm Lazy are used to reduce data transmission between clouds during the transition process,and finally reduce the transition cost.Through theoretical verification,ClouT can improve the storage efficiency to a certain extent and reduce the cost of storage and transition.The test results show:compared with the traditional transition process,using the intermediate node coding and the delay-based batch transition algorithm Lazy can make the cost optimization ratio up to62%,performance optimization ratio up to 27%,and achieve the expected performance and cost optimization results.