Research On Ant Colony Algorithms With Initial Pheromone Screening In HDFS Replica Selection

The demand for efficient storage of mass data storage has become increasingly stringent with the further application and development of information. How to store and manage these data which all show a tendency of exponential growth efficiently and safely has become a research focus in the huge amounts of data. The rapid development of distributed technology, makes the distributed a effective way to solve the storage and management of mass data. The data in the cloud storage system adopts a method of distributed storage. And, at the same time also have some replicas are saved to the different data node. This measure not only guarantees the security of the data, and improves the efficiency of the concurrent read from the same piece of data.The technology of replica is the key to ensure performance and reliability in cloud storage. And replica selcetion is the basis of data access and management in cloud storage, the merits of replica selection will affect directly the performance of the system, load balancing and reliability.Therefore, how to choose the best replica from several replicas in order to improve best the rate of access is an important of replica management. To slove the above problems, this paper puts forward a strategy of replica selection, which is based on the initial pheromone screening of ant colony algorithm. The main research contents of the paper are as follows:(1) In view of the existing replica selection strategy of the longger response time, the poorer load balancing problems, and based on the principle of the replica selection strategy research, put forward the major factors that affcct replica selection. They are disk I/O transmission rate, network bandwidth, the load balancing of replica nodes and the physical distance between the reauest node and replica node. In addition, I also analyse the necessity of replica selection in the environment of HDFS.(2) To compare and analyze two replica selection stagtegies based on ACO and GA, puts forward a model of replica selection in the environment of HDFS, which is based on the initial pheromone screening of ant colony algorithm. By using the genetic algorithm to gain several groups of optimization solution, the specific steps which could be used include coding, initializing the initial population, defining the fitness function, selection, crossover and mutation. And, the steps are a cycle operation. At the same time, the selection operation using an improved roulette wheel selection method to obtain a better population.This paper using a dynamic linking method in order to make the genetic algorithm link up with ant colony algorithm at the best time. Taking advantage of the optimization solution through genetic algorithm to initialize the pheromone, then using ant colony algorithm to obtain the best replica. Besides, to ensure load balancing of the replica nodes, combine the probability of choosing with the load completion of replica nodes.(3)The partial class of CloudSim is modified and extended, and adding the initial pheromone screening of ant colony algorithm, genetic algorithm and ant colony algorithm to CloudSim. Comparing the effect of three algorithms in terms of the job execution time, the replica's response time and the load of replica by the extended simulation platform.Simulation results show that the initial pheromone screening of ant colony algorithmcan gets the best of replica more quickly. In addition, it increases the load balancing of the system and enhances the overall performance of the system.