Performance-aware placement and chaining scheme for virtualized network functions: a particle swarm optimization approach

Abstract

Network functions virtualization (NFV) is a new concept that has received the attention of both researchers and network providers. NFV decouples network functions from specialized hardware devices and virtualizes these network functions as software instances called virtualized network functions (VNFs). NFV leads to various benefits, including more flexibility, high resource utilization, and easy upgrades and maintenances. Despite recent works in this field, placement and chaining of VNFs need more attention. More specifically, some of the existing works have considered only the placement of VNFs and ignored the chaining part. So, they have not provided an integrated view of host or bandwidth resources and propagation delay of paths. In this paper, we solve the VNF placement and chaining problem as an optimization problem based on the particle swarm optimization (PSO) algorithm. Our goal is to minimize the required number of used servers, the average propagation delay of paths, and the average utilization of links while meeting network demands and constraints. Based on the obtained results, the algorithm proposed in this study can find feasible and high-quality solutions.

Discrete spider monkey optimization algorithm for latency‐sensitive VNF deployment and resource allocation

“…The CrossoverFraction is 0.8, the MigrationFraction is 0.2, the PopulationSize and Generation is equal to DSMO. • PSO 48 : The particle swarm optimization algorithm simulates animals' behavior by designing a massless particle with only two attributes: position and velocity. Position represents the direction of movement and velocity represents the speed of movement.…”

“…The following schemes are the baselines for our simulation. Such as the genetic algorithm 47 (denoted as GA), the particle swarm optimization algorithm 48 (denoted as PSO), the firefly algorithm 49 (denoted as FA), the integer encoding grey wolf optimizer 10 (denoted as IEGWO) and the whale optimization algorithm 50 (denoted as WOA). Firstly, they are well‐known metaheuristics in solving NP‐hard problems.…”

“… GA 47 : The genetic algorithm is a heuristic algorithm used to search for an optimal solution by simulating the natural evolutionary process, such as selection, mutation and crossover. The CrossoverFraction is 0.8, the MigrationFraction is 0.2, the PopulationSize and Generation is equal to DSMO. PSO 48 : The particle swarm optimization algorithm simulates animals' behavior by designing a massless particle with only two attributes: position and velocity. Position represents the direction of movement and velocity represents the speed of movement.…”

In the Direction of Service Guarantees for Virtualized Network Functions

“…and use learning algorithms to make predictions about the performance of VNFs. The work presented in [28] suggests a profiling framework to accurately model the performance of VNF. Similarly, the work presented in [29] proposes a profiling-based solution that can model the performance of a network functions chain (multiple VNFS connected in a series).…”