Research On The Performance Optimization For Clouds-of-cloud Storage System Based On I/O Access Frequency

The storing pattern of data has been gradually influenced by the trend of data explosion.The appearance of cloud storage technology makes it easier for companies and users to store and manage data.In recent years,however,some noticeable events of which cloud storage services break off or shut down have brought about severe economic damages for those who simply store their data in one cloud.As a result,the data availability and security has been continuously attracting attention in the field of cloud storage.Cloud-of-cloud storage system distributes data separately into several different clouds.Redundancy mechanism is supposed to reconstitute the unavailable data and files when a certain cloud breaks down,which effectively tackles the problem arosen from depending on a single cloud such as vendor lock-in and low data availability.Meanwhile,the application of encryption technology and erasure code also ensure the security of data in each cloud.There have been plenty of researches on cloud-of-cloud storage system,but they failed to combine the access performance with storage cost to evaluate the cloud services.On the other hand,researches on the relationship between the workload characteristics and cloud-of-cloud performance usually focus on the data type and data size.After analyzing characteristics of different cloud services synthetically,this paper proposes two new standards,i.e.,performance-price ratio and storage-price ratio,to evaluate the cloud storage services.At the same time,after analyzing the workload characteristics,this paper proposes a high performance cloud-of-cloud storage system based on the data access frequency of workload.The performance-price ratio is proposed to measure the access performance of cloud service and the storage-price ratio is proposed to measure the storage expenditure.The cloud-of-cloud storage system designed by this paper stores those data which are frequently accessed into the clouds with high performance-price ratio so as to promote the performance with lower respond delay,and stores other data with less accessed frequency into the clouds with high storage-price ratio in order to eliminate the storage expenditure.