Design And Research For Application Layer Multicast Algorithm Based On Intelligent Algorithm

The application layer multicast technology came out with the rapid development of network technology. Due to the limitations of IP multicast on technology and market application, the application layer multicast appeared as an alternative technology and gradually stepped into the practical application. At some point to multipoint applications such as video conference, remote education, online live broadcast, video applications, etc, network bandwidth and delay have lots of limitation. Common network applications are based on unicast way, but this can’t deal with some of the above applications. The application layer multicast technique implements multicast in the application layer. This has great advantages in the actual deployment.In the study of the application layer multicast, multicast tree construction is a kind of important research subject. Data needs to be carried on multicast tree after arriving in the multicast group. So the construction of multicast tree has a very big impact on the performance of the multicast group. This thesis analyzes the factors which may affect multicast performance, and eventually uses constrained minimum spanning tree as the multicast tree topology. Because the constraint minimum spanning tree (DCMST)problem is a kind of NP-hard problem, it cannot be solved in polynomial time. People generally adopt a series of optimal approximation algorithms for solving this problem.A series of methods to solve DCMST problem is based on intelligent algorithms. We put forward two tree building algorithms:one is based on genetic algorithm and the other is based on particle swarm optimization algorithm and estimation of distribution algorithm. The experiment and simulation prove the feasibility of our algorithms.The constrained minimum spanning tree constructing algorithm based on genetic algorithm (GA-based DCMST, GA-DCMST) solves DCMST problem by simulating natural evolution. This algorithm estimates the optimal solution by operations of encoding, building, selecting, intersecting and variation, specified conditions are set to terminate the iteration. The algorithm finds the optimal solution in global scope, and has the feature of fast convergence speed, convergence of high efficiency, strong scalability and parallelism good characteristics.The proposed DCMST algorithm based on particle swarm optimization and estimation of distribution algorithm(PSO and EDA-based DCMST, PE-DCMST) combines advantages of particle swarm optimization (PSO) and estimation of distribution algorithm(EDA). PSO algorithm is based on the concept of the foraging behavior of birds. It makes the optimal vectors as a particle in space, and finds the optimal solution by evaluating particle’s movement. If the particles achieve stability or satisfy the specified conditions, it is an optimal solution. Our algorithm combines PSO and EDA and has the trait of good convergence results, fast speed, and the characteristics of parallel performance.Facing the problem of setting parameters in the tree building algorithm, this thesis proposes a parameter adaptive optimization method based on artificial fish swarm algorithm (AFSA-based Self-adaptive Parameter Adjusting, AFSA-SPA). This method is to adjust parameters of PE-DCMST method in order to get the best parameters of PSO algorithm. This algorithm is concise, efficient and has a fast convergence speed.Through experiments and simulation, we validate the rationality of GA-DCMST algorithm and PE-DCMST algorithm. In the process of experiment, the GA-DCMST algorithm shows good parallelism and has a fast convergence speed, but there are also some shortcomings, such as the complexity and difficulty of implementation, many parameter Settings and Possibility of bad results. In comparison, PE-DCMST algorithm has a better effect in parallelism and efficiency. The PE-DCMST algorithm in combination with AFSA-SPA algorithm is good to solve the problems of DCMST problem. Our research has certain guiding significance and practical significance on application layer multicast research, and obviously can be applied in real applications.