Research On Network And Workload Aware Cloud Storage Synchronization Technology

At present,the global digital society is becoming more and more mature.The use of the cloud storage has become so pervasive that many organizations adopts a "cloudfirst" approach.It is foreseeable that all organizations will use cloud services in the future.In the last six years,file sharing and file collaboration become the most used category of cloud services,which brings more serious challenges for file sharing and collaboration in cloud services,as well as file synchronization.In view of the challenges which cloud synchronization receives,how to improve the performance of cloud synchronization and provide high-quality services for a large number of users is an urgent problem to be solved.We firstly analyze the synchronization efficiency and performance problem of existing cloud synchronization technologies in large-scale cloud service application scenarios.Our experimental results show that delta synchronization schemes incur high processing overhead for both the client and server when synchronizing small files,while the bandwidth efficiency of full synchronization schemes is quite high.Expecially,the repeated data transmission of redundant data will induce long synchronization latency when synchonizing large files.Based on the characteristics of the workload and network latency,this paper proposes a hybrid network and workload aware cloud synchronization scheme,called PandaSync,which combines the strengths of both delta synchronization and full synchronization methods by dynamically and judiciously switching between them.Thus,PandaSync improves the performance and quality of service of the cloud synchronization.Moreover,to further improve small-file synchronization performance,we propose an optimization,called Full2Sync,which merges the synchronization requests with the file-sending requests to reduce the number of network round-trips between the client and the cloud servers.For the large-file synchronization,we propose an optimization,called DeltaSync(F&B FSC),which utilizes the backward fixed size chunking to further eliminate the redundant data.The experiments conducted on our lightweight prototype implementation of PandaSync show that PandaSync reduces the synchronization time by an average of 87.6%and 78.1%compared with the delta synchronization scheme and full synchronization scheme,respectively.In summary,the network and workload aware cloud storage synchronization Technology(PandaSync)can adapt to the changes of the workload characteristics and network latency,improve cloud synchronization performance significantly,provide better sensing services for users,and provide better consistency guarantees for cloud storage systems.