Genetic Algorithm for Optimizing the Virtual Network Functions Placement and Traffic Routing Problem.

Abstract

 Recently, the Internet infrastructure is constantly being improved to meet the increasing needs of users. Installing
hardware devices in the network server center faces many limitations such as: difficulty in replacing devices; large
investment and operating costs,... Network Function Virtualization (NFV) technology was invented to overcome the
disadvantages. Specialized hardware will be replaced by software running in visualized environments in this technology.
Resource allocation is an important problem in NFV. Reasonable resource allocation can increase service quality and
reduce network deployment costs. Optimizing resource allocation can be accomplished through placing virtual network
functions (VNFs) on servers and finding appropriate routing for network services. This research explores the network
resource allocation to achieve two objectives: i) reduce deployment costs; ii) reduce the delay of the service chain
in the network. We transfer the multi-objective problem to single-objective problem (SO-VPTR) by the weighted-sum
approach. This research proposes a mixed-integer linear programming (MILP) model to solve the single-objective
optimization problem. The MILP model can provide accurate solutions for small datasets. In order to find solutions
for larger sized datasets, the research proposes a genetic algorithm to solve the problem. The experimental results on
different datasets demonstrated the effectiveness of the proposed algorithm