The Research Of Data Storage In Distributed Main Memory Database

In recent years, the domestic and international researches on the Smart Grid vigorously promote the process of Grid information. With the deepening of the information technology on the Grid system, the power grid system requires more storage capacity and real-time ability on the database system, so the main memory database which has more powerful performance than the traditional disk database has been widely used in the power grid system. However, due to the limited main memory capacity, main memory database always appears as the role of cache system in the power grid system to alleviate the performance bottleneck of the traditional disk database.With the rise of the Smart Grid, compared to the traditional power grid system, the data interaction is more and more frequent and the amount of data is larger and larger. This requires the data platform of the Smart Grid should have greater data capacity and higher data concurrency. On the one hand means that the main memory database in the traditional grid system should be charge with more business data and business applications, it also sets the new requirements and challenges to the main memory database.With the characteristics of the grid and the trend of the Smart Grid development, the paper proposes the database system which is a distributed main memory database established on a distributed main memory file system. To solve the shortage of the memory capacity and concurrency in centralized main memory database system. The system is divided into two layers. The lower is the distributed memory file system, providing the memory file services to the upper layer, is responsible for the load balance and the data consistency in the all nodes of the system. The upper layer is the distributed memory database management system. By calling the services of the file system below, it stores the relational data and the index of all these data. The two-layer system is a loosely coupled architecture, each responsible for separate functions and optimization.In the single node condition, the TPS(Transaction Per Second) of the database is higher the disk 1 to 2 orders of magnitude; and in the system, each additional node can provide about 0.76 times the single-node concurrency(linear growth); when a node joins or leaves the system, all the nodes can be load balanced in a considerable short time.