Resource Allocation and Optimal Scheduling of Virtual Network Functions in Software Defined Networks

Publication Type:
Issue Date:
Full metadata record
One of the main challenges that faces the Network Functions Virtualization (NFV) deployment is to optimize the resource allocation of demanded network services in the NFV environment. In this study, new optimization models have been developed to find the near to optimal mapping and scheduling for the incoming Virtual Network Function (VNF) requests. The optimization models are formulated as a multi-objective problem in general where different objectives and constraints can be defined depending on the considered scenarios. In the first formulation, three objectives have been defined, namely, maximizing the number of accepted incoming service requests, optimizing link utilization and minimizing the overall processing time of service requests. The second development includes an optimization problem that considers the nonuniform arrival of the incoming service requests periodically. This optimization problem has been done by maximizing the number of accepted service requests, minimizing the number of bottleneck links, the overall processing time. In the third development, the optimization problem considers the expiry time for those incoming service requests to be processed in the VMs. Moreover, the model considers the uniform and non-uniform arrival of the incoming service requests. Four different objectives and five constraints have been considered to solve this optimization problem. Particularly, the model aims to maximize the acceptance rate, minimize the number of bottleneck links, the overall processing time and the relative processing time. In the fourth scenario, the optimization model has been developed to achieve three objectives functions, namely, minimizing the transmission delays occurring in every link, minimizing the processing capacity for every VM and minimizing the processing delay at every VM. The optimization model developed in the fifth formulation minimizes the processing time for every accepted service request, and at the same time maximizes the number of accepted service requests. All five scenarios have been treated as both single-objective and multi-objective optimization problems, where two different evolutionary algorithms based on a genetic algorithm have been applied for solving the resulting optimization problems. Via numerical simulations, it is shown that for the first three scenarios, the proposed algorithms solve the problem efficiently and converge to near to the optimal solution. Regarding the latter two scenarios, the numerical evaluations provide an evidence that the algorithms developed in this manuscript are scalable and they outperform the evolutionary algorithms proposed in the literature, namely genetic bandwidth link allocation (GA-BA) and genetic non-bandwidth link allocation (GA-NBA) algorithms.
Please use this identifier to cite or link to this item: