GRASP and ILS Based Meta-Heuristic Approximation Algorithms for Optimal Virtual Network Synthesis (VNS) in Multi-AS Environment

Emerging Internet architecture utilizes a pluralistic architectural paradigm in which multiple virtual overlay service networks are built on top of the existing interconnected Autonomous System (multi-AS) networks across the world to satisfy diverse and ever-demanding service requirements from end users as well as emerging network applications. However, optimal provisioning and management of virtual networks in the multi-AS environment is still a challenging and unsolved problem. The problem has been extensively studied within the research community as a so-called multi-domain virtual network embedding (MD-VNE) optimization problem, but the current research results so far have been unsatisfactory and lack practical use. This paper proposes a bottom-up virtual network synthesis (VNS) approach to the MD-VNE problem in multi-AS network environment and formulates it into a combinatorial optimization problem. We then propose two trajectory-based meta-heuristic approximation algorithms based on Greedy Randomized Adaptive Search Procedure (GRASP) and Iterative Local Search (ILS) methods together with a simple Greedy Heuristic (GH)-based algorithm as approximate solutions to the multi-AS VNS optimization problem. Experiments with these three approximation algorithms were conducted, and their performances were analyzed quantitatively based on well-defined metrics. The analysis results show the feasibility of the developed approximation solutions for practical solutions to the MD-VNE problem.