Development of dynamic resource management mechanisms for virtualized network environments

Abstract

The unprecedented and extreme cloud computing growth, virtualization of network functions, and software-defined networking paradigms have driven and guided many technological developments. These advances made it possible to significantly and somewhat radically review the design, deployment, and management of network services. Despite the promises in terms of reduced operating costs, automated operations, and increased revenue, several challenges remain to overcome to be able to make the transition to full automation of network and service management. Among the challenges, the placement and chaining of virtualized network functions and the dynamic adaptation of these remain among the most studied. In practice, these two problems are considered very promising for making this transition successful. Simultaneously, they are interconnected and require effective and practical solutions that meet the demands of the applications and services they are intended for when deployed and reconfigured to accommodate specific changes at a lower cost. In this thesis, the two problems are studied, and different techniques have been proposed to solve them. We have considered different scenarios for each research problem and evaluated using several performance metrics. Furthermore, it contains four contributions: First, FASTCALE, a scalable cultural genetic algorithm is proposed for the placement and chaining of complex virtual network services; Second, ARTIMIS a suite of optimization techniques for the selection of VNFs, given their different performances and configurations, for delay-sensitive services and a chemical reaction based meta-heuristic for their placement and chaining. Third, VALKYRIE a set of clustering techniques enabling the deployment of service function chains across on-demand clusters and the reduction of the search space toward guaranteed feasible solutions; Forth, DAVINCI a dynamic adaptation approach that incorporates elasticity mechanisms as a set of decisions to adapt the service function chains with least cost quickly. The evaluation results are presented to show the effectiveness of the proposed models and algorithms.

Date	15 Apr 2021
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Nadjia Kara (Supervisor)