Research On Efficient Replica Management Strategy In Cloud Environment

With the rapid development of Internet technology, especially the popularity of web2.0 applications, data begins to grow explosively. Although cloud storage can solve the bottleneckeffectivelythatthe traditional file system can handle the big data on the capacity and performance, it also tends to cause some problems such as the data is not available.Besides, cloud storage can’t realize load flexibly and effectively when the data access changes so that costshigh energy. So how tomake cloud storage provide efficient and high reliable data storage services and reduce energy consumption of the whole system effectively has become the top priority.At the present, the existing researches use multiple replicas management strategy to achieve load balancing in the system, but lack of taking energy efficient strategies into consideration.In addition, another deficiency is ubiquitous in existing energy efficient technology: thepresent implementation has been proportional to the energy consumption, but could not support the system load balancingperfectly, and few studies take full advantage of multiple replicas of data to achieve the target that energy efficiently.This thesis aims at those problems mentioned above onmultiple replicas in cloud storage technology research, studies about how to realize load balancing and energy efficient in the management strategy of copies, and gets through the management of multiple replicas to realize the data access efficiency and energy efficient. The main content is depicted as follows:(1) In the aspect of load balance: firstly, taking use of theimplementation principle of load balancing and the analysis of the tidal characteristics of file, the heatcalculation model offile access is proposed based on time series, and aduplicate factor allocation algorithm centered on the heat rankingof file access is given which is based on thecalculation model to achieve the goal that the replica number can beadjusted dynamically. Then this paper proposes a replica placement mechanismbased on theplacement price, which means it willchoose the best node to place a replicaaccording to the size of the placing price. Finally this paper puts forward the replica selection strategy based on weighted Euclidean distance, making full use of the advantages on the multiple attribute decision in solving problems, regarding the value of distance as an evaluation standard for replica selection, so that the optimal replica from multiple replicas can be chosen for users. Through the simulation experiments, itshows that theload variance in adaptive load balancing replicamanagement strategy decreases by 34% than static replica strategy, the average execution time of operation decreasesby 27%, which means the system performance is improvedobviously and it can provide users with efficient data storage service.(2) In terms of energy efficient: firstly, by means of the life cycle characteristics of files and the analysis of the characteristics of the data access and energy supply and demand, this paper reveals the causes of low energy efficiency. Then the efficient energy dynamic replica placement mechanism is proposed, in which combining the data features of multiple replicas combined with the specific replica placement layout to preparefor the realization of the scalable system. Finally this paper gives the cluster flex design based on the energy awareness to improve the energy efficiency according to the sleep data node and the wake up data node in load statement of system. Through the simulation experiment, it shows thatin achieving efficientenergy the efficient energy dynamic replica placement strategy reduced the energy consumptionby 16% than the static copy strategy.