Research On VNF Deployment And Its Parallelization Based On Meta-Heuristics

Traditional networks require the deployment of specific physical devices to provide tailor-made functionality to meet user service requests,which can lead to network hardening and cost too much for upgrades and maintenance.Network functions virtualization(NFV),a good solution to the above problems,achieves flexible and dynamic deployment and migration of network functions through hardware and software separation and virtualization technology.A key problem in the implementation of NFV is the deployment of VNF.How to find the optimal solution to deploy VNF on the underlying hardware under various resource constraints and costs is the core of this paper.Aiming at minimizing the end-to-end delay as the optimization objective,this paper proposes three improved meta-heuristic algorithms to solve the problem,and explores the use of Spark platform for parallelizing the algorithm to improve algorithm performance.The specific contents are as follows:1)This paper proposes an improved whale optimization algorithm(WOA).Inspired by genetic algorithm(GA)and fruit fly optimization algorithm(FOA),this algorithm combines the crossover operation and the mutation strategy,and designs an optimal individual guidance strategy based on the probability concentration to enhance its solution quality and convergence.Experimental results show that compared with the classical particle swarm optimization(PSO)and GA,this algorithm can get a better deployment scheme.2)This paper proposes an improved grey wolf optimizer(GWO).The algorithm designs an adaptive strategy for the individual and divides the individual dynamically into local search wolf or global search wolf according to the fitness.Global search wolves perform global search operations based on the incremental updates to increase diversity.Local search wolves implement a local mining strategy based on the ant colony pathfinding to enhance search results.The simulation results show that the algorithm can get the best results among all the algorithms,and the algorithm is relatively stable.3)This paper proposes an improved cuckoo search(CS).Inspired by PSO,the algorithm stores the historical optimal solution of searching till now in each individual,and introduces a global factor and a local factor to control individual updating,and designs two ways of logical neighborhood search and physical neighborhood search to strengthen the local mining.The results show that this algorithm can get a better solution than the traditional meta-heuristic algorithms,and the algorithm is faster.4)This paper proposes two parallel methods of meta-heuristic algorithm.Aiming at this kind of meta-heuristic algorithm cluster based on population search,this paper analyzes its characteristics in detail,proposes a parallelization model,and implements a general parallelization method based on Spark.On the basis of the general parallelization method,a customized and improved parallelization method was designed.The final results show that the improved parallel method has a better solution,can enhance the stability of the algorithm,and in large-scale scenarios or problems can speed up the algorithm execution efficiency,improving the speed of operation.