Research On Replication Technology In Cloud Storage For Smart Grid

In recent years, with the development of national smart grid, smart grid data increased exponentially. It not only includes structured data, but also includes some semi-structured and unstructured audio and video data. It has becoming an urgent problem for smart grid development to storage securely and processing effectively.Cloud storage system has the character such as high security, large storage capacity,easy expansion and so on, it can effectively store massive amounts of data that collected by smart grid. The replicationtion technology of cloud storage can effectively guarantee the security for data and the concurrency for processing tasks.However it has brought a series of questions such as replication consistency, load balance, replication access cost. And a good data placement strategy can solve the problem of load balancing, data selection policy can select the smallest cost of data replication. The replication placement strategy and replication selection policy in cloud storage for smart grid are the focuses in this paper.First, a replication placement strategy based on the relative storage capacity and a adjust strategy based on the heat of replication access were proposed.Beacause of the irregularity of data placement strategy in smart grid cloud storage system, according to the relative storage capacity of heterogeneous node to storage data, that can ensure the load balancing, and according the correlation of data, by setting the correlation coefficient, the relational data is stored centrally to ensure the relational query speed. Based on the dynamic demands for the number of replications for smart grid data processing tasks, according to the adaptive feedback mechanism, based on heat data access, increasing or decreasing the replication of the data to balance the read load and save the storage space.Secondly, a least cost of selection algorithm based on local data was proposed.When the smart grid processing tasks chose data replications, they did it randomly,without regard to access cost. This algorithm is based on the different access cost of data, choosing the least cost data replication in all the replications. In this paper, we also optimized the Map Reduce parallel programming model, according to the centralized storage for related data, does the data connection operating in Map stages, avoiding the data pull in Reduce phrase, reducing the transmission of irrelevant data. The data reading time is shorter 35% than the traditional cloud data storage system.This article embarks from the replication issue in cloud storage system,Closely around the two problems: replication placement strategy, replicationselection strategy,we do a in-depth research. A replication placement strategy based on relative storage capacity and a least cost of selection algorithm based on local data were proposed. The two schemes can effectively solve problems of replication placement and selection for smart grid.