Research On Network Coding Oriented Mulitcast Routing Optimization Problem

With the rapid development of information technologies,such as cloud computing,big data and so on,the Internet technologies,for example software-defined network and mobile Internet,are evolving promptly.It causes that the demands for multimedia applications increasing dramatically.Multicast technology can better support real-time,high-bandwidth multimedia services.However,traditional routing adopts the “store-and-forward” mode to transmit data,and it can not guarantee the maximum throughput of multicast.Different from traditional data transmission method,network coding employs "code-and-forward" mode to transmit data,which not only guarantees theoretical maximum throughput for multicast,and also has significant advantages in term of saving energy consumption and improving data security.Therefore,network coding has important theoretical value and great potential for application.However,the coding operation involves complex mathematical operations.Excessive coding operations will undoubtedly consume a lot of computing and storage resources,which increases in multicast costs.Therefore,under the condition of satisfying the multicast data rate,how to reduce the number of coding operations to reduce the overall network overhead is an extremely important academic research issue.This problem is called the Network Coding Resource Minimization(NCRM)problem.On the other hand,network service providers expect to maximize the utilization of network resources to save network costs.The load balancing is one of the most important indicators to measure the utilization of network resources.In the context of network coding based multicast,network coding brings theoretical maximum throughput,at the same time,balancing the network load as much as possible,which has become one of the important topics in the field of communication research.It has become one of the most important topics for future network research.This is called Load Balancing in Network Coding Based Multicast(LBNCM)problem.Compared with the traditional optimization algorithms,evolutionary algorithms have strong global search capability,simple implementation,and high robustness.It is widely used in engineering practice,especially in solving combinatorial optimization problems and NP-Hard problems.Therefore,evolutionary algorithms are used in this paper to solve NCRM and LBNCM problems.The specific implementation is described as follows:1)For the NCRM problem in static network environment,this paper proposes an improved particle swarm optimization algorithm to solve it.A greedy initialization strategy guides the algorithm quickly to the feasible solution field.The path relinking local search strategy is to further enhance the local search ability of PSO.The experimental results indicate that the proposed algorithm has high stability and less computational time,and it can obtain a multicast subgraph with less coding times for NCRM problem.2)This paper studys the NCRM problem in dynamic network environment and adopts a multiagent evolutionary algorithm to address it.The speed of information transmission between agents in MAEA is fast.When the network environment changes,each agent will quickly adapt to the changes in the environment.The Experimental results show that MAEA can attain fewer number of coding operation than genetic algorithm,quantum-inspired evolutionary algorithm,and population incremental learning algorithm before the network environment changes.3)Regarding the problem of LBNCM,this paper presents a modified artificial bee colony algorithm to tackle it.The algorithm integrates three new strategies.The guiding food source initialization strategy provides a food source with high nectar amounts within initial population.A nectar-source-library-based selection scheme further searches for high quality solutions.A probability distribution model search strategy strenghthens population diversity and avoids premature convergence.A number of simulation experiments on 12 fixed topologies and 12 randomly generated topologies demonstrate that the proposed algorithm can well balance the network load.