Model-driven Development of Virtual Network Embedding Algorithms with Model Transformation and Linear Optimization Techniques

Abstract

Enhancing the scalability and utilization of data centers, virtualization is a promising technology to manage, develop and operate network functions in a flexible way. For the placement of virtual networks in the data center, many approaches and algorithms are discussed in the literature to optimize solving the so-called virtual network embedding problem with respect to various optimization goals. This paper presents a new approach for the model-driven specification, simulation-based evaluation, and implementation of possible mapping algorithms that respect a set of given constraints and using linear optimization solving techniques to select one almost optimal mapping. Rule-based model transformation techniques are used to translate a given mapping problem into a linear optimization problem by taking domain specific knowledge into account. The resulting framework thus supports the design and evaluation of (correct-by-construction) virtual network embedding algorithms on a high level of abstraction. Well-defined model transformation rule refinement strategies can be used to reduce the search space for the employed linear optimization techniques.