Research On The Cloud Storage Load Balance Based On User Behavior From Server Side

With the fast development of technology, as well as the global data grows explosively, people need a new way to solve the problem of high cost of data and low reliability. Then cloud storage is formulated and acquires the attention of IT because of its' large capacity, high performance,low cost and high reliability etc. To improve the whole system performance and solve the problems of kinds of storage devices, load balance, system reliability and availability, researchers invented the Software Define Storage and replication technology. So we can use the software to mange the files' replication, including creation, replication, deletation,location,consistency maintenance to improve the availability and performance of the system. This thesis researches on the cloud storage load balance based on user behavior from the server side and use a software define cloud storage model to find effective ways to of the replication layout, the replication adjustment strategy. The main work of this thesis includes:1. Analyzes the research fruit of others about the cloud storage and key technology of load balance. Focuses on the replication management technology, this thesis summarizes its characteristics and problems.2. For the shortcomings of the cloud storage system and considering the new characteristics of cloud storage, this thesis proposes a Software Define Cloud Storage Based on User Behavior(USDCS). In the model, the server side controller can manage the storage points and the key data of documents. This paper focuses on the video storage which means reading operation is more often than writing.3. Considering the distribution ways, business property and high reliability request, this thesis proposes a replication strategy based on the user behavior and genetic algorithm (LUGS) . This strategy can use the relationship between user behavior and location to help choose the location of replication to improve the load balance and decline the visit cost of users for the documents in the cloud system.4. Focuses on the replication adjustment problem, this thesis proposes a replication adjustment strategy based on the file heat prediction (FHPRS ).This thesis uses an AR prediction model to predict the file heat and proposes the concept of load balance contribution, excessive degree of file heat, excessive degree of node's load. Finally we can calculate the number of the replication and use LUGS to find the replication place dynamically as well as reduce the cost user visit the system.