A significant fraction of data traffic is transmitted via cellular networks. When introducing fifth-generation (5G) radio access technology, the maximum bitrate of the radio link increases significantly, and the delay is lowered. Network congestion occurs when the sender attempts to send data at a higher rate than the network link or nodes can handle. In order to improve the performance of the mobile networks, many congestion control techniques and approaches have been developed over the years. Varying radio conditions in mobile networks make it challenging to indicate the occurrence of the congestion using packet loss as congestion indicator. This master thesis develops a congestion control algorithm based on Artificial Intelligence (AI) technologies, evaluates and compares it with existing state-of-the-art congestion control algorithms that are used with TCP today.In this study, we use the abundant readable physical layer information exchanged between the base stations and the user equipment to predict the available bandwidth. Two neural network models, Multi-Layer Perceptron (MLP) and Long Short-Term Memory (LSTM), are introduced as congestion control algorithms based on cross-layer information in order to improve user throughput and utilize the available capacity as much as possible.Evaluation in a Long-Term Evolution (LTE) network system simulator confirms that the estimation of LSTM model is able to track the varying link capacity, while MLP is less accurate and induces higher delay. The sender uses the estimated link capacity to adjust its packet sending behavior. Our evaluation reveals that for large flows, the LSTM model can attain higher throughput than state-of-the-art congestion control algorithms, which are the Google Bottleneck Bandwidth and Round-trip propagation time (BBR) algorithm and the Data Center TCP (DCTCP) algorithm. However, it has higher latency than that of these two algorithms. The MLP based model provides unstable performance compared to LSTM; its prediction is not accurate enough and has the highest latency among the algorithms.In conclusion, the LSTM does not underperform the state-of-the-art congestion control algorithms. However, it does not provide additional performance gains in current settings. The MLP model underperforms BBR and DCTCP with L4S and it is not stable enough to be used as a congestion control algorithms. / En betydande del av datatrafiken överförs via mobilnät. Vid introduktion av femte generationens (5G) radioåtkomstteknik ökar den maximala bithastigheten för radiolänken betydligt och förseningen sänks. Nätstockning uppstår när avsändaren försöker skicka data med högre hastighet än nätverkslänken eller noderna kan hantera. För att förbättra prestandan i mobilnät har många tekniker för trängselkontroll utvecklats under åren. Varierande radioförhållanden i mobilnätet gör det utmanande att indikera förekomsten av trängseln med hjälp av paketförlust som trängselindikator. Detta examensarbete utvecklar en trängselkontrollalgoritm baserad på AI-teknik (Artificial Intelligence), utvärderar och jämför den med befintliga toppmoderna trängselkontrollalgoritmer som används med TCP idag.I denna studie använder vi den rikliga läsbara informationen om fysiskt lager som utbyts mellan basstationerna och användarutrustningen för att förutsäga den tillgängliga bandbredden. Två neurala nätverksmodeller, Multi-Layer Perceptron (MLP) och Long Short-Term Memory (LSTM), introduceras som trängselkontrollalgoritmer baserade på tvärskiktsinformation för att förbättra användarens genomströmning och utnyttja den tillgängliga kapaciteten så mycket som möjligt.Utvärdering i en LTE-nätverkssystemsimulator (Long Term Evolution) bekräftar att uppskattningen av LSTM-modellen kan spåra den varierande länkkapaciteten, medan MLP är mindre exakt och inducerar högre fördröjning. Avsändaren använder den uppskattade länkkapaciteten för att justera sitt paketets sändningsbeteende. Vår utvärdering avslöjar att för stora flöden kan LSTM-modellen uppnå högre genomströmning än modernaste trängselkontrollalgoritmer, som är Google Bottleneck Bandbredd och BBR-algoritm och Data Center TCP (DCTCP) ) algoritm. Men det har högre latens än för dessa två algoritmer. Den MLP-baserade modellen ger instabil prestanda jämfört med LSTM; dess förutsägelse är inte nog noggrann och har den högsta latensen bland algoritmerna.Sammanfattningsvis underpresterar LSTM inte de senaste toppkontrollalgoritmerna. Det ger emellertid inte ytterligare prestationsvinster i de aktuella inställningarna. MLP-modellen underpresterar BBR och DCTCP med L4S och den är inte tillräckligt stabil för att användas som en överbelastningskontrollalgoritm.

Active Queue Management (AQM) schemes are used for ensuring the Quality of Service (QoS) in telecommunication networks. However, they are sensitive to parameter settings and have weaknesses in detecting and controlling congestion under dynamically changing network situations. Another drawback for the AQM algorithms is that they have been applied only on the Markovian models which are considered as Short Range Dependent (SRD) traffic models. However, traffic measurements from communication networks have shown that network traffic can exhibit self-similar as well as Long Range Dependent (LRD) properties. Therefore, it is important to design new algorithms not only to control congestion but also to have the ability to predict the onset of congestion within a network. An aim of this research is to devise some new congestion control methods for communication networks that make use of various traffic characteristics, such as LRD, which has not previously been employed in congestion control methods currently used in the Internet. A queueing model with a number of ON/OFF sources has been used and this incorporates a novel congestion prediction algorithm for AQM. The simulation results have shown that applying the algorithm can provide better performance than an equivalent system without the prediction. Modifying the algorithm by the inclusion of a sliding window mechanism has been shown to further improve the performance in terms of controlling the total number of packets within the system and improving the throughput. Also considered is the important problem of maintaining QoS constraints, such as mean delay, which is crucially important in providing satisfactory transmission of real-time services over multi-service networks like the Internet and which were not originally designed for this purpose. An algorithm has been developed to provide a control strategy that operates on a buffer which incorporates a moveable threshold. The algorithm has been developed to control the mean delay by dynamically adjusting the threshold, which, in turn, controls the effective arrival rate by randomly dropping packets. This work has been carried out using a mixture of computer simulation and analytical modelling. The performance of the new methods that have.

Cette thèse analyse la possibilité d'utiliser des communications sans fil inter-véhiculaires pour améliorer la sécurité routière. Les performances du nouveau réseau ainsi créé (réseau ad-hoc véhiculaire) sont étudiées analytiquement et par des simulations dans un environnement réaliste. La thèse se concentre surtout sur des scénarios avec une forte densité de véhicules. Dans ce cas, l'accès au support devient un problème essentiel, en principal pour les applications de sécurité routière qui nécessitent une qualité de service élevée pour fonctionner dans un tel contexte. Ce travail montre que la version actuelle du standard IEEE 802.11, proposé comme méthode d'accès dans les réseaux véhiculaires, ne peut pas résoudre ce problème de passage à l'échelle pour supporter correctement les applications de sécurité routière. Plusieurs améliorations possibles sont analysées, liées à l'utilisation optimale de certains paramètres du protocole comme la taille de la fenêtre de contention ou bien le seuil de détection de la porteuse. Des nouveaux mécanismes adaptatifs visant ces paramètres sont proposés et les améliorations ainsi obtenues sont non-négligeables. Finalement, une nouvelle méthode d'accès est définie, en tenant compte des caractéristiques des applications de sécurité routière. Toujours basée sur des techniques CSMA, cette technique donne des résultats largement supérieurs à la version standard actuelle. / The equipment of vehicles with wireless communication devices in order to improve road safety is a major component of a future intelligent transportation system. The success and availability of IEEE 802.11-based products make this technology the main competitor for the Medium Access Control (MAC) layer used in vehicle-to-vehicle communication. The IEEE 802.11p amendment has been specially designed in this special context of wireless access in vehicular environments. However, as all the other approaches based on Carrier Sense Multiple Access (CSMA), this protocol presents scalability problems, which leads to poor performance in high density scenarios, quite frequent in the case of a vehicular ad hoc network (VANET). This thesis studies the congestion control problem in the context of safety vehicular communications, with a special focus on the back-off mechanism and the carrier sense function. First of all, a number of important characteristics presented by the safety messages are discovered and understood by the means of an analytical framework. Second, the lessons learned from the analytical study are put into practice with the design of two adaptive mechanisms (one for the contention window and the other one for the carrier sense threshold) that take into account the local vehicular density. These mechanisms remain simple, but highly efficient, while also being straightforward to integrate in IEEE 802.11 devices. Finally, by taking into account the most important properties of a safety VANET, a new CSMA-based MAC protocol is proposed. This new access method, named Safety Range CSMA (SR-CSMA), relies on the idea that collisions can not be avoided in a high density network. However, by increasing the number of simultaneous transmissions between geographically distant nodes, SR-CSMA manages to better protect the immediate neighborhood, the most important area for safety applications.

Since the field of computer networks has rapidly grown in the last two decades, congestion control of traffic loads within networks has become a high priority. Congestion occurs in network routers when the number of incoming packets exceeds the available network resources, such as buffer space and bandwidth allocation. This may result in a poor network performance with reference to average packet queueing delay, packet loss rate and throughput. To enhance the performance when the network becomes congested, several different active queue management (AQM) methods have been proposed and some of these are discussed in this thesis. Specifically, these AQM methods are surveyed in detail and their strengths and limitations are highlighted. A comparison is conducted between five known AQM methods, Random Early Detection (RED), Gentle Random Early Detection (GRED), Adaptive Random Early Detection (ARED), Dynamic Random Early Drop (DRED) and BLUE, based on several performance measures, including mean queue length, throughput, average queueing delay, overflow packet loss probability, packet dropping probability and the total of overflow loss and dropping probabilities for packets, with the aim of identifying which AQM method gives the most satisfactory results of the performance measures. This thesis presents a new AQM approach based on the RED algorithm that determines and controls the congested router buffers in an early stage. This approach is called Dynamic RED (REDD), which stabilises the average queue length between minimum and maximum threshold positions at a certain level called the target level to prevent building up the queues in the router buffers. A comparison is made between the proposed REDD, RED and ARED approaches regarding the above performance measures. Moreover, three methods based on RED and fuzzy logic are proposed to control the congested router buffers incipiently. These methods are named REDD1, REDD2, and REDD3 and their performances are also compared with RED using the above performance measures to identify which method achieves the most satisfactory results. Furthermore, a set of discrete-time queue analytical models are developed based on the following approaches: RED, GRED, DRED and BLUE, to detect the congestion at router buffers in an early stage. The proposed analytical models use the instantaneous queue length as a congestion measure to capture short term changes in the input and prevent packet loss due to overflow. The proposed analytical models are experimentally compared with their corresponding AQM simulations with reference to the above performance measures to identify which approach gives the most satisfactory results. The simulations for RED, GRED, ARED, DRED, BLUE, REDD, REDD1, REDD2 and REDD3 are run ten times, each time with a change of seed and the results of each run are used to obtain mean values, variance, standard deviation and 95% confidence intervals. The performance measures are calculated based on data collected only after the system has reached a steady state. After extensive experimentation, the results show that the proposed REDD, REDD1, REDD2 and REDD3 algorithms and some of the proposed analytical models such as DRED-Alpha, RED and GRED models offer somewhat better results of mean queue length and average queueing delay than these achieved by RED and its variants when the values of packet arrival probability are greater than the value of packet departure probability, i.e. in a congestion situation. This suggests that when traffic is largely of a non bursty nature, instantaneous queue length might be a better congestion measure to use rather than the average queue length as in the more traditional models.

Lors d'une communication dans un réseau, les nœuds intermédiaires se contentent en général de retransmettre les paquets de données qu'ils reçoivent. Grâce au codage de réseau (NC), ces nœuds intermédiaires peuvent envoyer des combinaisons linéaires des paquets qu'ils ont reçus. Ceci permet une meilleure exploitation de la capacité du réseau et une plus grande robustesse à l'égard de pertes.Cette thèse s'intéresse à une implantation du NC en lien avec TCP (TCP-NC). Grâce à la redondance introduite par le NC, une partie des pertes liées à des liens sans fils peut être compensée. Elle propose en particulier un mécanisme d'adaptation de la redondance introduite par le codage de réseau. Une première partie de cette thèse est consacrée à l'analyse de la dynamique de TCP-NC avec Random Early Detection (RED) comme mécanisme de gestion des files d'attente en utilisant les outils d'optimisation convexe et issus de l’automatique. Nous caractérisons l'équilibre du réseau et les propriétés de stabilité de TCP-Reno en présence de NC. Dans une seconde partie, cette thèse propose un algorithme d'adaptation de la redondance introduite par NC. Dans TCP-NC avec redondance adaptative (TCP-NCAR), cet ajustement se fait grâce à un schéma de différenciation des pertes, qui estime la répartition des pertes entre erreurs de transmission dues aux liens sans fils et pertes liées à la congestion. Les propriétés d'équilibre et de stabilité de TCP-NCAR/RED sont caractérisées. Les résultats théoriques et de simulation montrent que TCP-NCAR adopte une redondance proche de l'optimum quand les taux de perte de paquets sur les liens sans fils sont petits. En outre, le modèle linéarisé autour de l'équilibre montre que TCP-NCAR augmente la taille de la région de stabilité de TCP-Reno. / Communication networks today share the same fundamental principle of operation: information is delivered to their destination by nodes intermediate in a store-and-forward manner.Network coding (NC) is a technique that allows intermediate nodes to send out packets that are linear combinations of previously received information. The main benefits of NC are the potential throughput improvements and a high degree of robustness, which is translated into loss resilience. These benefits have motivated deployment efforts for practical applications of NC, e.g., incorporating NC into congestion control schemes such as TCP-Reno to get a TCP-NC congestion protocol. In TCP-NC, TCP-Reno throughput is improved by sending a fixed amount of redundant packets, which mask part of the losses due, e.g., to channel transmission errors. In this thesis, we first analyze the dynamics of TCP-NC with random early detection (RED) as active queue management (AQM) using tools from convex optimization and feedback control. We study the network equilibrium point and the stability properties of TCP-Reno when NC is incorporated into the TCP/IP protocol stack. The existence and uniqueness of an equilibrium point is proved, and characterized in terms of average throughput, loss rate, and queue length. Our study also shows that TCP-NC/RED becomes unstable when delay or link capacities increases, but also, when the amount of redundant packets added by NC increases. Using a continuous-time model and neglecting feedback delays, we prove that TCP-NC is globally stable. We provide a sufficient condition for local stability when feedback delays are present. The fairness of TCP-NC with respect to TCP-Reno-like protocols is also studied. Second, we propose an algorithm to dynamically adjust the amount of redundant linear combinations of packets transmitted by NC. In TCP-NC with adaptive redundancy (TCP-NCAR), the redundancy is adjusted using a loss differentiation scheme, which estimates the amount of losses due to channel transmission errors and due to congestion. Simulation results show that TCP-NCAR outperforms TCP-NC in terms of throughput. Finally, we analyze the equilibrium and stability properties of TCP-NCAR/RED. The existence and uniqueness of an equilibrium point is characterized experimentally. The TCP-NCAR/RED dynamics are modeled using a continuous-time model. Theoretical and simulation results show that TCP-NCAR tracks the optimal value for the redundancy for small values of the packet loss rate. Moreover, simulations of the linearized model around equilibrium show that TCP-NCAR increases the size of the TCP-Reno stability region. We show that this is due to the compensator effect of the redundancy adaptation dynamics to TCP-Reno. These characteristics of TCP-NCAR allow the congestion window adaptation mechanism of TCP-Reno to react in a smooth way to channel losses, avoiding some unnecessary rate reductions, and increasing the local stability of TCP-Reno.

Performanz ist ein zentraler Bestandteil des Designs von Anonymisierungsdiensten. Ihre zunehmende Popularität führt jedoch zu einer hohen Netzwerklast, die unzulängliche Entwurfsentscheidungen imminent macht. Die Anforderungen und die vielschichtige Architektur von Anonymisierungsdiensten machen die Thematik zu einem anspruchsvollen und zugleich inspirierenden Forschungsgegenstand. Die vorliegende Arbeit diskutiert das Design von sogenannten Niedriglatenz-Anonymisierungsdiensten im Allgemeinen und dem Tor-Netzwerk als relevantesten Vertreter im Speziellen. Es werden Lösungen für eine Reihe von Forschungsfragen entwickelt, die allesamt das Ziel verfolgen, diese Overlay-Netzwerke zu verbessern und sicherer zu gestalten. Es entsteht ein fundamentales Verständnis zu Netzwerkaspekten in Anonymisierungs-Overlays, das die Netzwerklast, als vorherrschende Ursache für die schwache Performanz, thematisiert. / Performance is a pivot point in the design of anonymity overlays. Due to their growing popularity, they are faced with increasing load, which makes design problems imminent. The special requirements and complex architecture of anonymity overlays renders the topic a challenging but likewise inspiring object of research. In this work, we discuss the design of low-latency anonymous communication systems in general and the Tor network as the de-facto standard in particular. We develop solutions to a number of research questions, all collectively following the aim of enhancing and securing such networks. By doing this we create a fundamental technical understanding of networking aspects in anonymity overlays and tackle the most prevalent performance issue experienced today: network congestion.

"The Internet has assumed the role of the underlying communication network for applications such as file transfer, electronic mail, Web browsing and multimedia streaming. Multimedia streaming, in particular, is growing with the growth in power and connectivity of today's computers. These Internet applications have a variety of network service requirements and traffic characteristics, which presents new challenges to the single best-effort service of today's Internet. TCP, the de facto Internet transport protocol, has been successful in satisfying the needs of traditional Internet applications, but fails to satisfy the increasingly popular delay sensitive multimedia applications. Streaming applications often use UDP without a proper congestion avoidance mechanisms, threatening the well-being of the Internet. This dissertation presents an IP router traffic management mechanism, referred to as Crimson, that can be seamlessly deployed in the current Internet to protect well-behaving traffic from misbehaving traffic and support Quality of Service (QoS) requirements of delay sensitive multimedia applications as well as traditional Internet applications. In addition, as a means to enhance Internet support for multimedia streaming, this dissertation report presents design and evaluation of a TCP-Friendly and streaming-friendly transport protocol called the Multimedia Transport Protocol (MTP). Through a simulation study this report shows the Crimson network efficiently handles network congestion and minimizes queuing delay while providing affordable fairness protection from misbehaving flows over a wide range of traffic conditions. In addition, our results show that MTP offers streaming performance comparable to that provided by UDP, while doing so under a TCP-Friendly rate."

Les exigences de Qualité de Service (QoS) des applications VANET varient selon la nature et le type de l’application. Par conséquent, un protocole de communication VANET doit pouvoir répondre aux diverses exigences de QoS selon le type du trafic. Dans VANET, le canal de transmission est partagé par tous les véhicules en utilisant une même fréquence radio. Une mauvaise exploitation du canal peut donc conduire à des collisions et peut aussi engendrer un gaspillage de la bande passante. Un protocole MAC doit être alors conçu pour partager le canal entre les différents noeuds d’une manière efficace et équitable.Dans cette thèse nous présentons les contributions suivantes :1- Analyse et amélioration de la diffusion dans la norme IEEE 802.11.2- Optimisation de la technique CSMA pour des réseaux 1D et 2D.3- Développement d’un algorithme CSMA de transmission adaptatif qui met à jour le taux de détection de la porteuse en fonction d’une valeur de référence.4- Étude du gain obtenu par l’utilisation d’antennes directionnelles pour Aloha, Aloha non-slotté, et CSMA. / Quality of Service (QoS) requirements for VANET applications vary depending on the nature and type of the application. Therefore, a communication protocol in VANETs must be able to meet various QoS requirements according to the type of traffic. In VANET, the transmission channel is shared by all the vehicles using the same radio frequency. A poor exploitation of the channel can therefore lead to collisions and wasted bandwidth. A MAC protocol must therefore be designed to share the channel between the different nodes in an efficient and fair way.In this thesis we present the following contributions:1- Analysis and improvement of diffusion in the IEEE 802.11 standard.2- Optimization of the CSMA technique for 1D and 2D networks.3- Design of an adaptive transmission algorithm that updates the Carrier Sense threshold to reach a target value.4- Study the gain obtained by the use of directional antennas for Aloha, non-slotted Aloha, and CSMA.

近年來，一個具有壅塞控制機制的傳輸協議DCCP被提出，期能取代UDP成為不可靠傳輸的主流協議。我們以NS-2網路模擬器和實際網路進行實驗，發現DCCP無法與其他傳輸協議公平分享頻寬，因此現行DCCP的設計，尚無法完全取代UDP。此外，目前DCCP以調整封包間隔的方式進行壅塞控制，也不適用於講求時效性的網路服務。 本研究首先以實驗證明，當網路情況不佳時，DCCP無法與其他傳輸協議公平的分享頻寬；當使用DCCP傳輸越洋長距離網路電話，如遇頻寬不足時，會因頻寬競爭力較弱而無法維持通話品質。本研究提出可變速率方法(Flexible Bit-rate)調整時效性網路服務的封包大小來進行壅塞控制，在維持一定服務品質之前提下，促進網路的和諧。我們在一個實際網路的實驗環境中評估以UDP、DCCP及可變速率三種方式傳輸網路電話封包的效能，結果顯示透過可變速率方法，能有效降低網路電話的封包遺失率，維持通話品質。 / With congestion-control ability, Datagram Congestion Control Protocol (DCCP) is expected to replace UDP as a mainstream unreliable transport protocol. But our study found that DCCP is not able to get a fair share of bandwidth under the competition of others transport protocols no matter in NS-2 simulation or real world networking environments. Furthermore, any congestion control protocol that postpones the transmission of packets may not be adequate to support time-sensitive network services. To maintain the quality of time-sensitive network services as well as to be TCP-friendly when facing network bandwidth fluctuation, we propose a Flexible Bit-rate congestion control mechanism for VoIP to adjust their data rate. Our experiments show that Flexible Bit-rate congestion control method could effectively reduce the packet loss rate and to maintain VoIP quality as compared with UDP and DCCP. Furthermore, it can have a much better bandwidth efficiency and adjust better to network fluctuation.

Load control design is one of the major cornerstones of radio resource management in today's UMTS networks. A WCDMA cell's ability to utilize available spectrum efficiently, maintain system stability and deliver minimum quality of service (QoS) requirements to in-cell users builds on the algorithms employed to manage the load. Admission control (AC) and congestion control (CC) are the two foremost techniques used for regulating the load, and differing environments will place varying requirements on the AC and CC schemes to optimize the QoS for the entire radio network. This thesis studies a real-life situation where cells are put under strenuous conditions, investigates the degrading effects a high-speed train has on the cell's ability to maintain acceptable levels of QoS, and proposes methods for mitigating these effects. The scenario is studied with regard to voice traffic where the limiting radio resource is downlink power. CC schemes that take levels of fairness into account between on-board train users and outdoor users are proposed and evaluated through simulation. Methods to anticipatorily adapt radio resource management (RRM) in a cell to prepare for a train is proposed and evaluated through simulation. A method to detect a high-speed train in a cell, and the users on it, is outlined and motivated but not simulated. Simulation results are promising but not conclusive. The suggested CC schemes show a surprising tendency towards an increase in congestion avoidance performance. Proactive RRM shows a significant increase in QoS for on-board users. No negative effects to users in the macro environment is noticed, with regard to the studied metrics.