In response to the presence of portable devices and the change in usage patterns, mobility support for computer networking has received increasing attentions in recent years. There has been wide variety of established works focusing on the shortening of the handover delay. Most of them had tried their best in maintaining the connections to their peer nodes as mobile nodes moving across different network domains. This might not be a best policy in certain applications, such as homogeneous content distribution systems. A new handover scheme incorporating anycasting and service migration is developed and presented in this thesis. In the proposed approach, anycast is used to locate a new nearest server as a mobile node entering a new domain, and an ongoing service will be dynamically migrated from a distant server to a new server with shorter weighted network distance to the client. As a result, the individual service connection, as well as the global network environment, will benefit from the service migration, in terms of improved service quality and bandwidth utilization.

With increased popularity and pervasiveness, mobile networking had become a definite trend for future networks. Users strongly demand the retaining of the connectivity and the QoS (Quality of Service) of ongoing services while roaming across different points of attachment. Efficient handover schemes and service paradigms are essential to the above vision. We will contribute to the QoS provisioning in mobile wireless networks from two complementary perspectives: one is the improved handover schemes at the client end for shorter latency and less packet loss, and the other is the service migration at server end for improved QoS. There are time-consuming procedures involved in the handover process. Various research works had devoted to the acceleration of movement detection and registration. However, a time-consuming operation, duplicate address detection, was overlooked by most studies. A novel scheme featuring anycast / multicast technique is developed and presented in this dissertation. The proposed approach switches to anycast / multicast addressing during handover and switches back to normal unicast addressing after all required operations are completed. By switching to anycast / multicast addressing, a mobile node can continue the reception of packets from its corresponding node before its new care-of address is actually validated. As a result, transmission disruption can be effectively minimized. In addition, simple but effective buffer control schemes are designed to reduce possible packet loss and to prevent the out-of-order problem. Analytical study reveals that improved performance can be guaranteed, as reflected in the simulation results. The establishment of mobility-supported Internet protocols, such as IPv4 and IPv6, had made it possible that an ongoing service can be retained while a mobile node is roaming across different access domains. However, limited efforts had been paid to server sides if we consider the topological change due to node mobility. In the global network environment, the weighted network distance between a client and its server could change dramatically for reasons of topology change or node mobility. A new network service framework highlighting the concept of service migration is presented in this dissertation. The proposed framework take into account essential service quality factors, such as server loading, bandwidth, delay, and so on, and then dynamically migrates an ongoing service from a distant server to a new server with shorter ¡§weighted network distance¡¨ to the client. As a result, the individual service connection, as well as the global network environment, will benefit from the service migration, in terms of improved service quality and bandwidth utilization. This dissertation explains the general architecture of the proposed framework and focuses on the technical details of the core component - service migration module. Our experiences on the functional prototypes for service migration are also reported. The success of the prototyping system is an indication of the feasibility and effectiveness of the proposed scheme.

Over the past few years, Deep Learning (DL), a promising technology leading the next generation of intelligent environments, has attracted significant attention and has been intensively utilized in various fields in the fourth industrial revolution era. The applications of Deep Learning in the area of Mobile Edge Computing (MEC) have achieved remarkable outcomes. Among several functionalities of MEC, the service migration frameworks have been proposed to overcome the shortcomings of the traditional methodologies in supporting high-mobility users with real-time responses. The service migration in MEC is a complex optimization problem that considers several dynamic environmental factors to make an optimal decision on whether, when, and where to migrate. In line with the trend, various service migration frameworks based on a variety of optimization algorithms have been proposed to overcome the limitations of the traditional methodologies. However, it is required to devise a more sophisticated and realistic model by solving the computational complexity and improving the inefficiency of existing frameworks. Therefore, an efficient service migration mechanism that is able to capture the environmental variables comprehensively is required. In this thesis, we propose an enhanced service migration model to address user proximity issues. We first introduce innovative service migration models for single-user and multi-user to overcome the users’ proximity issue while enforcing the service execution efficiency. Secondly, We formulate the service migration process as a complicated optimization problem and utilize Deep Reinforcement Learning (DRL) to estimate the optimal policy to minimize the migration cost, transaction cost, and consumed energy jointly. Lastly, we compare the proposed models with existing migration methodologies through analytical simulations from various aspects. The numerical results demonstrate that the proposed models can estimate the optimal policy despite the computational complexity caused by the dynamic environment and high-mobility users.

none

Les travaux décrits dans cette thèse de doctorat visent à traiter les futures architectures de fourniture de services mobiles basés sur le cloud, à travers l'évolution des infrastructures réseau partant de Mobile Cloud Computing (MCC) au Mobile Edge Computing (MEC). Nous nous sommes essentiellement concentrés sur le concept Follow Me Cloud (FMC) comme une nouvelle stratégie de fourniture de services pour une meilleure expérience utilisateur et une utilisation efficace des ressources. Cela permet aux services basés sur le cloud de "suivre" leurs utilisateurs mobiles au cours de leurs déplacements à travers les technologies de réseau d'accès, tout en fournissant le service basé sur le cloud via le point de service le plus optimal au sein de l'infrastructure cloud. Plusieurs contributions sont proposées dans cette thèse, avec des évaluations à la fois en analyse théorique et en simulation scientifique.Premièrement, nous avons proposé une architecture alternative FMC qui permet: (i) d'ouvrir la conception FMC sur les technologies d'accès réseau mobile non-3GPP (ii) d'assurer l'interopérabilité entre différents domaines PMIPv6 permettant au MN une itinérance inter-domaines PMIPv6 avec une mobilité IP transparente ainsi qu'une continuité de session de service.(iii) d'offrir une architecture sans tunnel dans les situations d'itinérance de MN, en évitant ainsi toute surcharge supplémentaire liée aux tunnels dans la gestion de la mobilité. Le schéma proposé exploite la technologie SDN/OpenFlow et le protocole de gestion de la mobilité PMIPv6 en les intégrant dans un unique framework permettant de réaliser la vision FMC.Deuxièmement, pour aborder les problèmes d'évolutivité et de résilience dans les architectures SDN/OpenFlow centralisées de plan de contrôle, nous avons introduit une nouvelle conception d'un contrôleur SDN élastique et distribué adapté pour MCC et plus particulièrement pour les systèmes de gestion FMC. Nous avons illustré comment le nouveau schéma de plan de contrôle est distribué sur une architecture hiérarchique à deux niveaux, un premier niveau avec un seul contrôleur SDN global et un second niveau avec plusieurs contrôleurs SDN locaux. Ensuite, nous avons présenté les éléments constitutifs de notre nouvel framework de plan contrôle, le calcul de l'indicateur de performance (KPI) du système, et nous avons fixé l'objectif clé de notre conception visant à maintenir la valeur KPI du système dans une fenêtre de seuil prédéfinie. Enfin, nous avons démontré comment cet objectif est atteint en adaptant dynamiquement le nombre et l'emplacement des contrôleurs SDN locaux en utilisant la technologie NFV pour provisionner les contrôleurs SDN en tant que instances VNF (fonction réseau virtuelle) dans le cloud.Troisièmement, nous avons introduit le concept FMeC, exploitant les capacités offertes par la combinaison des architectures MEC et FMC dans le but de satisfaire aux exigences des systèmes automobiles 5G. Nous avons commencé par définir les éléments clés du concept FMeC permettant de fournir la technologie FMC en bordure des réseaux mobiles. Ensuite, nous avons présenté une projection de notre solution FMeC sur un cas d'utilisation de conduite automatisée intégrant l'industrie automobile aux infrastructures Telecom en vue de la vision automobile 5G future. Avec une focalisation sur les types de communications V2I/N, nous avons présenté la conception de notre architecture FMeC basée sur les technologies SDN/OpenFlow et les entités de l'infrastructure MEC dont les ressources sont mises en commun pour fournir un cloud de bordure fédéré. Enfin, nous avons présenté notre framework sensible à la mobilité pour le placement des services dans le cloud de bordure, ce dernier est fondé sur un ensemble d'algorithmes de base qui permettent d'atteindre les exigences de QoS de la conduite automatisée en termes de latence ultra-courte au sein du réseau 5G. / This Ph.D. thesis aims to deal with the future delivery architectures of mobile cloud-based services, through network infrastructures evolving from Mobile Cloud Computing (MCC) to Mobile Edge Computing (MEC). We mainly focused on Follow Me Cloud (FMC) concept as a new service delivery strategy for improved user experience and efficient resource utilization. That enables cloud-based services to follow their mobile users during their movement across access network technologies and by delivering the cloud-service via the optimal service point inside the cloud infrastructure. Several contributions are proposed in this thesis and evaluated in both theoretical analysis and scientific simulation.First, we proposed an alternative FMC architecture that allows: (i) to open the FMC design on non-3GPP mobile network access technologies (ii) to provide interoperability among different PMIPv6 domains permitting MNs inter-PMIPv6 domain roaming with seamless IP mobility and service session continuity (iii) to offer a tunnel-free architecture in MNs roaming situation, avoiding any additional overhead associated with tunneling in mobility management. This proposed scheme leverage SDN/OpenFlow technology and PMIPv6 mobility management protocol by integrating them within a framework permitting to realize the FMC vision.Second, to address the scalability and resiliency concerns in centralized SDN/OpenFlow control plane architecture, we introduced a new design of an elastic distributed SDN controller tailored for Mobile Cloud Computing (MCC) and more notably for Follow Me Cloud (FMC) management systems. We illustrated how the new control plane scheme is distributed on two-level hierarchical architecture, a first level with a single global SDN controller and a second level with several local SDN controllers. Then, we presented the building blocks of our novel control plane framework, the system Key Performance Indicator (KPI) computation and set the key objective of our design aiming to keep the system KPI value within a predefined threshold window. Last, we proved how this goal is achieved by adapting the number of local SDN controllers and their locations in an elastic manner and deploying them as VNF instances on the cloud thanks to NFV technology.Third, we introduced FMeC concept, leveraging the intertwining of MEC and FMC architectures with the aim of sustaining requirements of the 5G automotive systems. We began by defining FMeC key concept elements permitting to provide FMC technology at the edge of mobile networks. Then, we presented an automated driving use case projection of our FMeC solution integrating automotive with Telco infrastructures towards the future 5G automotive vision. Focusing on the V2I/N communications types, we introduced our FMeC design architecture based on SDN/OpenFlow technologies and MEC infrastructure entities whose resources are pooled together to provide a federated edge clouds. Finally, we presented our mobility-aware framework for edge-cloud service placement based on a set of basic algorithms that permit achieving the automated driving QoS requirements in terms of ultra-short latency within 5G network.