Research Of Replica Selection Strategy Based In Ant Algorithm In Data Grid

With the rapid development of computer application of information technology and people’s living standards improved gradually, we entered the area of information explosion. There is an unprecedented request for storage and the data reading speed. Besides that, with the size of the data storage increasing, the resource management will be more complex, and the challenges for scalability and reliability of storage will be increasingly difficult. In order to solve this series of problems we faced, data grid emerged. Data grid plays a role as "Super Computer" to integrate the mass dispersion, independent and heterogeneous data resources into a reliable and secure logical whole. According to the unified management of data, data grid provides users with transparent, reliable and efficient data services.In data grid, data can be saved in different nodes as the replicas. And then users can access the nearest replica or access the local replica directly. This approach reduces the access latency on the one hand, it also balances the load on the server, thus improving the availability of data greatly. In replica management system, one of the most important technology is replica selection. The technology is to choose the best replica from multiple replicas and to make the minimum cost.As we all know, replica selection in data grid is an optimization problem itself. And in recent years, computational intelligence algorithms have been widely used in optimization problems and achieve better results. Therefore, this paper mainly studies the improved algorithm based on ant colony algorithm and its application for replica selections in data grid. The main work includes:(1) This paper describes the basic principles of ant colony algorithm and the basic workflow of traditional ant colony algorithm. After introducing the experiment of simulating ants foraging process, we lead to the concept of pheromones.At the same time, we analysis several typical improved strategies of ant algorithm. We can know that the best algorithm is MMAS after comparing the experimental results. This paper also introduces the basic principle of the genetic algorithm. According to the characteristic of ant algorithm and genetic algorithm we can find their advantages and disadvantages, and we can blend the two algorithms to improve the performance of optimization.(2) On the basic of ant algorithm, we introduce the limits of pheromone adsorption mechanism. At last, we proposes the new algorithm——NGACO. Experimental results show that NGACO has the fastest convergence rate and the highest accuracy for optimization.(3) This paper presents a new strategy for replica selection based in NGACO in data grid. We make the simulation experiments according to a secondary development of Optorsim simulator. The results show that NGACO can reduce the average job execution time effectively, thus improving the overall performance of the replica management system.