Anycast Routing Of Qos Constraints Based On Evolutionary Algorithms

Anycast is a new standard communication method defined in IPv6, users can access the"best"one from a group of servers which identified by only one anycast address. The distribution of network traffic can be effectively balanced and the availability of services can be enhanced by anycast which has broad application space. Meanwhile, with the rapid prevalence speed of Internet and the continued development of network technologies, people are putting forward higher requirements for Quality of Service (QoS) in networks, as a result how to fix the routing problems with multi QoS constraints of anycast has became urgent.The major work and innovations in this thesis are as follows:(1) An approach to initialize population based on the load of member server in an anycast group is proposed. By this approach, the initial population is composed of various routes that from the source host to each server of destination group rather than to a fixed server, and the number of routes to a certain server is inversely proportional to the server's load. Therefore, the shortages of traditional algorithms in which the load of anycast group servers is not taken into account can be overcame, and the diversity of initial population can be highly increased, which is helpful to avoid trapping into local optimal and balance the network traffic as well as increasing the throughput of network.(2) A penalty function of QoS constraints with higher differentiation is proposed. Penalty function of QoS constraints for qualified routes is simply set in the traditional evolutionary algorithms for solving the anycast routing problems with QoS constraints, which is not effective to distinct which one is the best among those routes. But this shortage can be overcame by the penalty function of QoS constraints with higher differentiation, thus a route with better QoS performance can be find out, and the global optimal can be obtained rather than the local solutions.(3) For solving the problem of load non-balancing of anycast severs group, an anycast routing algorithm for QoS network load balancing based on evolutionary algorithm (QLBE) is proposed, an approach for generating the initial population based on the anycast group is introduced in QLBE, while the fitness function consist of a penalty function with higher differentiation for QoS constraints. Simulated experiments are carried out in network topologies randomly generated by Waxman's model. The results of experiments show that the best anycast route obtained by QLBE can effectively balance the load of anycast group servers in comparison with traditional algorithms, the iterations of convergence is fewer, in addition, the distribution of network traffic can be improved and the throughput of network can be increased.(4) For solving the problem of optimizing the QoS performance of anycast route, an anycast routing algorithm for the best QoS performance based on elite migration strategy of evolutionary algorithm (BQEM) is proposed, in which the path search space is divided into sub-populations by the size of anycast group, then the elite transfer strategy is introduced as the each sub-population is independently evolving. The elite individuals can be transferred among sub-populations in BQEM, it's beneficial to play the guiding role of elite individuals. The strategy is helpful to increase the diversity of sub-populations, avoid the blindfold evolving, and enhance the abilities of getting the global solutions and speed up the convergence. Eventually in the simulated experiments, the results demonstrate that delay and bandwidth performance of the anycast route obtained by BQEM is better than that of route obtained by EAs algorithms.