Virtual network function placement using Particle Swarm Optimization
Samane Asgari · 2020
Research Aim: Network functions (e.g., firewalls and network address translators) play an essential role in today's networks, as they support a diverse set of functions ranging from security (e.g., intrusion detection) to performance (e.g., proxy). Network functions are difficult to deploy and maintain and lead to high Capital expenditure and Operational expenditure. Network function virtualization is a new trend that transforms network functions into simple software examples referred to as virtual network functions. The network function virtualization has the advantages of enabling companies and organizations to reduce the cost of purchasing proprietary hardware, providing significant flexibility and leading to more efficient use of resources. One of the most challenging questions in network function virtualization is placement and chaining of virtual functions, because a good placement and chaining extensively effects on network performane. Research method: Since the problem of placement and chaining of virtual network functions belongs to complexity class NP-complete, so this problem could be solved using one of the integer linear programming, heuristic and meta heuristic strategies, and since integer programming is particularly time-consuming (several hours) especially for large-scale networks, heuristic algorithm may caught in the local optimal, meta-heuristic algorithms have been used, and among the meta-heuristic algorithms, considering the advantages such as finite number of parameters, simple calculations, easy implementation and less dependency on a set of initial points that the particle swarm optimization algorithm has compared to other techniques, this algorithm is used. Findings: Extensive simulations carried out in Net2Plan software show that the use of particle swarm optimization algorithm in placement and chaining of virtual network functions, improves network performance because it minimizes the number of servers used to host virtual functions, average link utilization and propagation delay of selected paths, algorithm execution time is very short, it has high acceptance rates, reduces power consumption and is scalable. Conclusion: By using particle swarm optimization algorithm, a good placement and chaining of virtual network functions can be reached, which increase the network performance.