當大型災害來臨，通訊系統對救災效益具有不可或缺的重要性。然而，一般公眾通訊系統，如行動通訊網路等，常因各種不同因素導致系統攤瘓，使得協調大批非組織的救災志工，顯得異常困難。現存多個緊急通訊系統的佈建，需仰頼良好的交通運輸。不幸的是，部分道路和橋樑常因大型災害而斷裂或變型，導致災區對外交通運輸中斷，無法快速的將緊急通訊系統的網路元件，運送至災區佈建。我們提出應急行動通訊系統(Contingency Cellular Network, CCN) 以提供災區救災工作的緊急通訊。部分行動基地台雖然結構完整，但因失去與核心網路連線能力或電力供應，而無法提供服務，成為孤立基地台。應急行動通訊網路(CCN) 搭配無線通訊與衛星通訊技術建置一多重跳接無線網路，以恢復孤立基地台與核心網路連線能力；並配備發電機，提供電力，使孤立基地台可提供有限的服務。救災志工和災民無需使用特殊手持設備或額外的訓練，只需使用原有的手機，即可使用CCN的應急通訊服務。CCN可於第一時間，提供大批救災志工和災民通訊服務，以提高救災效益，因而拯救更多寶貴的生命。 本論文主要聚焦在應急行動通訊系統設計所衍生出的相關議題，如 應急網路需求分析、系統架構設計、網路拓樸規劃、網路頻寬規劃、佈署行程規劃等議題。本論文針對網路拓樸規劃、網路頻寬規劃、佈署行程規劃問題以數學模式進行塑模並證明這些問題為NP-Hard問題。因網路拓樸規劃、網路頻寬規劃、佈署行程規劃需緊急完成，我們也提出啟發式算法快速解決這些規劃問題。實驗結果顯示，這些啟發式算法均具良好的效能。 / Communication system is crucial to the efficiency of disaster response operation in the large-scale disaster. However, communication systems, such as cellular networks, usually crashed due to various causes making coordination among disorganized disaster responders extremely difficult. Unfortunately, rapid deployment of many existing emergency communication systems relies on a good transportation system, which is usually not available in a catastrophic natural disaster. We proposed a Contingency Cellular Network (CCN) for emergency communication by connecting disconnected base stations together with wireless links to construct a wireless multi-hop cellular network. CCN can support existing mobile phone users with reduced capability. Such a system can support a large number of disaster responders in the early hours of a catastrophic natural disaster, thus save many lives. Our research addresses the design issues of the network topology of CCN, such as network topology planning, bandwidth management, deployment scheduling and etc., and we take the degree of emergency and population of each stricken area as the priority measure as well as the available resources as the constraint to determine the network topology. Mathematical models of these design issues are proposed and proved as NP-Hard problems. Since the network topology, bandwidth management, deployment scheduling are needed in urgent, we propose heuristic algorithms to solve these problems quickly. Finally, we evaluated the proposed algorithms by simulation. A significant improvement in resiliency is reached.

La gestió de xarxes és un camp molt ampli i inclou molts aspectes diferents. Aquesta tesi doctoral està centrada en la gestió dels recursos en les xarxes de banda ampla que disposin de mecanismes per fer reserves de recursos, com per exemple Asynchronous Transfer Mode (ATM) o Multi-Protocol Label Switching (MPLS). Es poden establir xarxes lògiques utilitzant els Virtual Paths (VP) d'ATM o els Label Switched Paths (LSP) de MPLS, als que anomenem genèricament camins lògics. Els usuaris de la xarxa utilitzen doncs aquests camins lògics, que poden tenir recursos assignats, per establir les seves comunicacions. A més, els camins lògics són molt flexibles i les seves característiques es poden canviar dinàmicament. Aquest treball, se centra, en particular, en la gestió dinàmica d'aquesta xarxa lògica per tal de maximitzar-ne el rendiment i adaptar-la a les connexions ofertes.En aquest escenari, hi ha diversos mecanismes que poden afectar i modificar les característiques dels camins lògics (ample de banda, ruta, etc.). Aquests mecanismes inclouen els de balanceig de la càrrega (reassignació d'ample de banda i reencaminament) i els de restauració de fallades (ús de camins lògics de backup). Aquests dos mecanismes poden modificar la xarxa lògica i gestionar els recursos (ample de banda) dels enllaços físics. Per tant, existeix la necessitat de coordinar aquests mecanismes per evitar possibles interferències. La gestió de recursos convencional que fa ús de la xarxa lògica, recalcula periòdicament (per exemple cada hora o cada dia) tota la xarxa lògica d'una forma centralitzada. Això introdueix el problema que els reajustaments de la xarxa lògica no es realitzen en el moment en què realment hi ha problemes. D'altra banda també introdueix la necessitat de mantenir una visió centralitzada de tota la xarxa. En aquesta tesi, es proposa una arquitectura distribuïda basada en un sistema multi agent. L'objectiu principal d'aquesta arquitectura és realitzar de forma conjunta i coordinada la gestió de recursos a nivell de xarxa lògica, integrant els mecanismes de reajustament d'ample de banda amb els mecanismes de restauració preplanejada, inclosa la gestió de l'ample de banda reservada per a la restauració. Es proposa que aquesta gestió es porti a terme d'una forma contínua, no periòdica, actuant quan es detecta el problema (quan un camí lògic està congestionat, o sigui, quan està rebutjant peticions de connexió dels usuaris perquè està saturat) i d'una forma completament distribuïda, o sigui, sense mantenir una visió global de la xarxa. Així doncs, l'arquitectura proposada realitza petits rearranjaments a la xarxa lògica adaptant-la d'una forma contínua a la demanda dels usuaris. L'arquitectura proposada també té en consideració altres objectius com l'escalabilitat, la modularitat, la robustesa, la flexibilitat i la simplicitat.El sistema multi agent proposat està estructurat en dues capes d'agents: els agents de monitorització (M) i els de rendiment (P). Aquests agents estan situats en els diferents nodes de la xarxa: hi ha un agent P i diversos agents M a cada node; aquests últims subordinats als P. Per tant l'arquitectura proposada es pot veure com una jerarquia d'agents. Cada agent és responsable de monitoritzar i controlar els recursos als que està assignat. S'han realitzat diferents experiments utilitzant un simulador distribuït a nivell de connexió proposat per nosaltres mateixos. Els resultats mostren que l'arquitectura proposada és capaç de realitzar les tasques assignades de detecció de la congestió, reassignació dinàmica d'ample de banda i reencaminament d'una forma coordinada amb els mecanismes de restauració preplanejada i gestió de l'ample de banda reservat per la restauració. L'arquitectura distribuïda ofereix una escalabilitat i robustesa acceptables gràcies a la seva flexibilitat i modularitat. / Network management is a wide field including many different topics. This thesis focuses on resource management of broadband networks that have the mechanisms for performing resource reservation, such as Asynchronous Transfer Mode (ATM) or Multi-Protocol Label Switching (MPLS). Logical networks can be established by using Virtual Paths (VP) in ATM or Label Switched Paths (LSP) in MPLS, which we call generically Logical Paths (LP). The network users then use these LPs, which can have pre-reserved resources, to establish their communications. Moreover, LPs are very flexible and their characteristics can be dynamically changed. This work focuses, in particular, on the dynamic management of these logical paths in order to maximise the network performance by adapting the logical network to the offered connections.In this scenario, there are several mechanisms that can affect and modify certain features of the LPs (bandwidth, route, etc.). They include load balancing mechanisms (bandwidth reallocation and re-routing) and fault restoration (utilisation of backup LPs). These two mechanisms can modify the logical network and manage the resources (bandwidth) of the physical links. Therefore, due to possible interferences, there is a need to co-ordinate these mechanisms. Conventional resource management, using a logical network, performs a centralised recalculation of the whole logical network periodically (e.g. every hour / day). This brings the problem that the logical network readjustments do not happen when a problem occurs. Moreover, there is a need of maintaining a centralised network overview. In this thesis, a distributed architecture, based on a Multi-Agent System (MAS), is proposed. The main objective of this architecture is to perform joint resource management at a logical network level, integrating the bandwidth reallocation and LP re-routing with pre-planned restoration and spare bandwidth management. This is performed continuously, not periodically, when a problem is detected (an LP is congested, i.e. it is rejecting new user connections because it is already saturated with user connections) in a completely distributed way, i.e. without any central network overview. Therefore, the proposed architecture performs small rearrangements in the logical network and thus it is continuously being adapted to the user demands. The proposed architecture also considers other objectives, such as scalability, modularity, robustness, simplicity and flexibility. The proposed MAS is structured in two layers of agents: The network Monitoring (M) agents and the Performance (P) agents. All these agents are situated at different network nodes, where the computing facilities are. There is one P agent and several M agents on every node. The M agents are subordinated to the P agents, therefore the proposed architecture can be seen as a hierarchy of agents. Each agent is responsible for monitoring and controlling the resources they are assigned to. We have performed several experiments, using a connection level distributed simulator of our own design. The results show that our architecture is capable of performing the assigned tasks of detecting congestion, dynamic bandwidth reallocation and re-routing in a co-ordinated way with the pre-planned restoration and the spare capacity management. The distributed architecture offers a suitable scalability and robustness due to its flexibility and modularity.

Le streaming adaptatif sur HTTP, désigné par HAS, est une technique de streaming vidéo largement déployée sur Internet. Elle utilise TCP comme protocole de transport. Elle consiste à segmenter la vidéo stockée sur un serveur web en petits segments indépendants de même durée de lecture et transcodés à plusieurs niveaux de qualité, désignés par "chunks". Le player, du côté du client HAS, demande périodiquement un nouveau chunk. Il sélectionne le niveau de qualité en se basant sur l’estimation de la bande passante du/des chunk(s) précédent(s). Étant donné que chaque client HAS est situé au sein d’un réseau d’accès, notre étude traite un cas fréquent dans l’usage quotidien: lorsque plusieurs clients partagent le même lien présentant un goulot d’étrangement et se trouvent en compétition sur la bande passante. Dans ce cas, on signale une dégradation de la qualité d’expérience (QoE) des utilisateurs de HAS et de la qualité de service (QoS) du réseau d’accès. Ainsi, l’objectif de cette thèse est d’optimiser le protocole TCP pour résoudre ces dégradations de QoE et QoS. Notre première contribution propose une méthode de bridage du débit HAS au niveau de la passerelle. Cette méthode est désignée par "Receive Window Tuning Method" (RWTM): elle utilise le principe de contrôle de flux de TCP et l’estimation passive du RTT au niveau de la passerelle. Nous avons comparé les performances de RWTM avec une méthode récente implémentée à la passerelle qui utilise une discipline particulière de gestion de la file d’attente, qui est désignée par "Hierarchical Token Bucket shaping Method" (HTBM). Les résultats d’évaluations indiquent que RWTM offre une meilleure QoE et une meilleure QoS de réseau d’accès que HTBM. Notre deuxième contribution consiste à mener une étude comparative combinant deux méthodes de bridages, RWTM et HTBM, avec quatre variantes TCP largement déployées, NewReno, Vegas, Illinois et Cubic. Les résultats d'évaluations montrent une discordance importante entre les performances des différentes combinaisons. De plus, la combinaison qui améliore les performances dans la majorité des scénarios étudiés est celle de RWTM avec Illinois. En outre, une mise à jour efficace de la valeur du paramètre "Slow Start Threshold", sthresh, peut accélérer la vitesse de convergence du player vers la qualité optimale. Notre troisième contribution propose une nouvelle variante de TCP adaptée aux flux HAS, qu’on désigne par TcpHas; c’est un algorithme de contrôle de congestion de TCP adapté aux spécifications de HAS. TcpHas estime le niveau de la qualité optimale du flux HAS en se basant sur l’estimation de la bande passante de bout en bout. Ensuite, TcpHas applique un bridage au trafic HAS en fonction du débit d’encodage du niveau de qualité estimé. TcpHas met à jour ssthresh pour accélérer la vitesse de convergence. Une étude comparative a été réalisée avec la variante Westwood+. Les résultats d’évaluations montrent que TcpHas est plus performant que Westwood+. / HTTP adaptive streaming (HAS) is a streaming video technique widely used over the Internet. It employs Transmission Control Protocol (TCP) as transport protocol and it splits the original video inside the server into segments of same duration, called "chunks", that are transcoded into multiple quality levels. The HAS player, on the client side, requests for one chunk each chunk duration and it commonly selects the quality level based on the estimated bandwidth of the previous chunk(s). Given that the HAS clients are located inside access networks, our investigation involves several HAS clients sharing the same bottleneck link and competing for bandwidth. Here, a degradation of both Quality of Experience (QoE) of HAS users and Quality of Service (QoS) of the access network are often recorded. The objective of this thesis is to optimize the TCP protocol in order to solve both QoE and QoS degradations. Our first contribution consists of proposing a gateway-based shaping method, that we called Receive Window Tuning Method (RWTM); it employs the TCP flow control and passive round trip time estimation on the gateway side. We compared the performances of RWTM with another gateway-based shaping method that is based on queuing discipline, called Hierarchical Token Bucket shaping Method (HTBM). The results of evaluation indicate that RWTM outperforms HTBM not only in terms of QoE of HAS but also in terms of QoS of access network by reducing the queuing delay and significantly reducing packet drop rate at the bottleneck.Our second contribution consists of a comparative evaluation when combining two shaping methods, RWTM and HTBM, and four very common TCP variants, NewReno, Vegas, Illinois and Cubic. The results show that there is a significant discordance in performance between combinations. Furthermore, the best combination that improves performances in the majority of scenarios is when combining Illinois variant with RWTM. In addition, the results reveal the importance of an efficient updating of the slow start threshold value, ssthresh, to accelerate the convergence toward the best feasible quality level. Our third contribution consists of proposing a novel HAS-based TCP variant, that we called TcpHas; it is a TCP congestion control algorithm that takes into consideration the specifications of HAS flow. Besides, it estimates the optimal quality level of its corresponding HAS flow based on end-to-end bandwidth estimation. Then, it permanently performs HAS traffic shaping based on the encoding rate of the estimated level. It also updates ssthresh to accelerate convergence speed. A comparative performance evaluation of TcpHas with a recent and well-known TCP variant that employs adaptive decrease mechanism, called Westwood+, was performed. Results indicated that TcpHas largely outperforms Westwood+; it offers better quality level stability on the optimal quality level, it dramatically reduces the packet drop rate and it generates lower queuing delay.