"Increasing availability of high-speed Internet and wireless access points has allowed home users to connect not only their computers but various other devices to the Internet. Every device running different applications requires unique Quality of Service (QoS). It has been shown that delay- sensitive applications, such as VoIP, remote login and online game sessions, suffer increased latency in the presence of throughput-sensitive applications such as FTP and P2P. Currently, there is no mechanism at the wireless AP to mitigate these effects except explicitly classifying the traffic based on port numbers or host IP addresses. We propose CHAP, a credit-based queue management technique, to eliminate the explicit configuration process and dynamically adjust the priority of all the flows from different devices to match their QoS requirements and wireless conditions to improve application quality in home networks. An analytical model is used to analyze the interaction between flows and credits and resulting queueing delays for packets. CHAP is evaluated using Network Simulator (NS2) under a wide range of conditions against First-In-First- Out (FIFO) and Strict Priority Queue (SPQ) scheduling algorithms. CHAP improves the quality of an online game, a VoIP session, a video streaming session, and a Web browsing activity by 20%, 3%, 93%, and 51%, respectively, compared to FIFO in the presence of an FTP download. CHAP provides these improvements similar to SPQ without an explicit classification of flows and a pre- configured scheduling policy. A Linux implementation of CHAP is used to evaluate its performance in a real residential network against FIFO. CHAP reduces the web response time by up to 85% compared to FIFO in the presence of a bulk file download. Our contributions include an analytic model for the credit-based queue management, simulation, and implementation of CHAP, which provides QoS with minimal configuration at the AP."

<p>Streaming video and VoIP are two popular services used over the Internet, and as the number of users increases, the demand on the network routers also increases. Since both streaming video and VoIP have a variable traffic flow, the routers must always have some free space in their receive buffers to handle traffic bursts. If not, packet loss may occur that will result in a degraded quality of the services. In this project, a fuzzy logic based Active Queue Management (AQM) will be implemented, which might help reduce this problem. This algorithm has currently only been tested in a simulated environment. This algorithm will then be evaluated and compared to some of the existing AQMs. The results will also be compared to a stream that only uses a First-In, First-Out (FIFO) queue, which will work as a baseline. Since an AQM is not the only means used to reduce delay and jitter, different AQMs will also be combined with two different transport protocols, User Datagram Protocol (UDP) and the new Datagram Congestion Control Protocol (DCCP). The resulting implementation performed on a par with RED (Random Early Detection), one of the most common AQMs.</p>

Achieving high performance in high capacity data transfers over the Internet has long been a daunting challenge. The current standard of Transmission Control Protocol (TCP), TCP Reno, does not scale efficiently to higher bandwidths. Various congestion controllers have been proposed to alleviate this problem. Most of these controllers primarily use marking/loss or/and delay as distinct feedback signals from the network, and employ separate data transfer control strategies that react to either marking/loss or delay. While these controllers have achieved better performance compared to existing TCP standard, they suffer from various shortcomings. Thus, in our previous work, we designed a congestion control scheme that jointly exploits both delay and marking; D+M (Delay Marking) TCP. We demonstrated that D+M TCP can adapt to highly dynamic network conditions and infrastructure using ns-2 simulations. Yet, an analytical explanation of D+M TCP was needed to explain why it works as observed. Furthermore, D+M TCP needed extensive simulations in order to assess its performance, especially in relation to other high-speed protocols. Therefore, we propose a model for D+M TCP based on distributed resource optimization theory. Based on this model, we argue that D+M TCP solves the network resource allocation problem in an optimal manner. Moreover, we analyze the fairness properties of D+M TCP, and its coexistence with different queue management algorithms. Resource optimization interpretation of D+M TCP allows us to derive equilibrium values of steady state of the controller, and we use ns-2 simulations to verify that the protocol indeed attains the analytical equilibria. Furthermore, dynamics of D+M TCP is also explained in a mathematical framework, and we show that D+M TCP achieves analytical predictions. Modeling the dynamics gives insights to the stability and convergence properties of D+M TCP, as we outline in the thesis. Moreover, we demonstrate that D+M TCP is able to achieve excellent performance in a variety of network conditions and infrastructure. D+M TCP achieved performance superior to most of the existing high-speed TCP versions in terms of link utilization, RTT fairness, goodput, and oscillatory behavior, as confirmed by comparative ns-2 simulations.

In this dissertation we examine some of the challenges faced by the congestion control algorithms of TCP in emerging networks. We focus on three main issues. First, we propose TCP with delayed congestion response (TCP-DCR), for improving performance in the presence of non-congestion events. TCP-DCR delays the conges- tion response for a short interval of time, allowing local recovery mechanisms to handle the event, if possible. If at the end of the delay, the event persists, it is treated as congestion loss. We evaluate TCP-DCR through analysis and simulations. Results show significant performance improvements in the presence of non-congestion events with marginal impact in their absence. TCP-DCR maintains fairness with standard TCP variants that respond immediately. Second, we propose Layered TCP (LTCP), which modifies a TCP flow to behave as a collection of virtual flows (or layers), to improve eficiency in high-speed networks. The number of layers is determined by dynamic network conditions. Convergence properties and RTT-unfairness are maintained similar to that of TCP. We provide the intuition and the design for the LTCP protocol and evaluation results based on both simulations and Linux implementation. Results show that LTCP is about an order of magnitude faster than TCP in utilizing high bandwidth links while maintaining promising convergence properties. Third, we study the feasibility of employing congestion avoidance algorithms in TCP. We show that end-host based congestion prediction is more accurate than previously characterized. However, uncertainties in congestion prediction may be un- avoidable. To address these uncertainties, we propose an end-host based mechanism called Probabilistic Early Response TCP (PERT). PERT emulates the probabilistic response function of the router-based scheme RED/ECN in the congestion response function of the end-host. We show through extensive simulations that, similar to router-based RED/ECN, PERT provides fair bandwidth sharing with low queuing delays and negligible packet losses, without requiring the router support. It exhibits better characteristics than TCP-Vegas, the illustrative end-host scheme. PERT can also be used for emulating other router schemes. We illustrate this through prelim- inary results for emulating the router-based mechanism REM/ECN. Finally, we show the interactions and benefits of combining the different proposed mechanisms.

Streaming video and VoIP are two popular services used over the Internet, and as the number of users increases, the demand on the network routers also increases. Since both streaming video and VoIP have a variable traffic flow, the routers must always have some free space in their receive buffers to handle traffic bursts. If not, packet loss may occur that will result in a degraded quality of the services. In this project, a fuzzy logic based Active Queue Management (AQM) will be implemented, which might help reduce this problem. This algorithm has currently only been tested in a simulated environment. This algorithm will then be evaluated and compared to some of the existing AQMs. The results will also be compared to a stream that only uses a First-In, First-Out (FIFO) queue, which will work as a baseline. Since an AQM is not the only means used to reduce delay and jitter, different AQMs will also be combined with two different transport protocols, User Datagram Protocol (UDP) and the new Datagram Congestion Control Protocol (DCCP). The resulting implementation performed on a par with RED (Random Early Detection), one of the most common AQMs.

La paralysie cérébrale (PC) cause souvent d’importants troubles de la marche. Suite à un examen clinique et une analyse quantifiée de la marche (AQM), ces troubles peuvent être traités par une chirurgie orthopédique, dite multi-sites, au cours de laquelle plusieurs corrections chirurgicales sont faites simultanément à différents niveaux des membres inférieurs. Les améliorations cinématiques apportées par ce traitement, bien que parfois très efficaces, demeurent à ce jour difficilement prévisibles. L’objectif de cette thèse est de simuler par apprentissage statistique supervisé l’effet de la chirurgie sur les signaux de marche, notamment les signaux cinématiques.Ce simulateur vise à montrer le résultat probable de la marche postopératoire afin d’aider à la décision chirurgicale. Une base de données constituée de 134 enfants atteints de PC, ayant été opérés et ayant eu au moins une AQM avant et après la chirurgie, a été exploitée. Les signaux cinématiques ont été prétraités et les données cliniques manquantes ont été imputées.Des caractéristiques des données prétraitées ont été extraites en utilisant différentes méthodes telles que l’approximation des courbes, la sélection de variables et la réduction de dimension par analyse en composantes principales. Ensuite des régressions ont été faites en utilisant différentes méthodes telles que la régression multilinéaire, la régression non linéaire avec des réseaux de neurones et l’apprentissage par ensembles. Les différentes méthodes testées ont été comparées entre elles, ainsi qu’avec d’autres méthodes trouvées dans la littérature. Il s’agit de la première fois que l’effet de la chirurgie sur la marche paralysée cérébrale est simulé de façon quantitative pour des nombreuses combinaisons chirurgicales et des nombreux patterns de marche. / Cerebral Palsy frequently leads to gait troubles. After a physical examination and a Clinical Gait Analysis (CGA), these walking troubles are usually treated by orthopedic surgery, called single event multi-level surgery (SEMLS), in which several surgical corrections are simultaneously done at different levels of the lower limbs. Kinematic improvements obtained by this treatment are sometimes very efficient, but at this moment they remain difficultly predictable. The objective of this thesis is to simulate the effect of surgery on gait parameters, using supervised statisticalmachine learning. The purpose of the simulator is to show the most likely gait outcome in order to improve decision-making in SEMLS. The database was composed of 134 children with cerebral palsy that have undergone surgery and have had at least one CGA before and after the treatment. Gait signals were preprocessed and physical examination missing data were imputed. Features of the preprocessed data were extracted using different techniques such ascurve fitting, variable selection and dimensionality reduction. Then regressions were performed utilizing different methods such as multiple linear regression, feedforward neural networks and ensemble learning. The tested methods and their performances were compared between them andto other methods in the literature. This work represents the first time that the effect of surgery on cerebral palsy gait is quantitatively simulated for a large number of surgical combinations and numerous different gait patterns.

The Internet, traditionally FTP, e-mail and Web traffic, is increasingly supporting emerging applications such as IP telephony, video conferencing and online games. These new genres of applications have different requirements in terms of throughput and delay than traditional applications. For example, interactive multimedia applications, unlike traditional applications, have more stringent delay constraints and less stringent loss constraints. Unfortunately, the current Internet offers a monolithic best-effort service to all applications without considering their specific requirements. Adaptive RED (ARED) is an Active Queue Management (AQM) technique, which optimizes the router for throughput. Throughput optimization provides acceptable QoS for traditional throughput sensitive applications, but is unfair for these new delay sensitive applications. While previous work has used different classes of QoS at the router to accommodate applications with varying requirements, thus far all have provided just 2 or 3 classes of service for applications to choose from. We propose two AQM mechanisms to optimize router for better overall QoS. Our first mechanism, RED-Worcester, is a simple extension to ARED in order to tune ARED for better average QoS support. Our second mechanism, REDBoston, further extends RED-Worcester to improve the QoS for all flows. Unlike earlier approaches, we do not predefine classes of service, but instead provide a continuum from which applications can choose. We evaluate our approach using NS-2 and present results showing the amount of improvement in QoS achieved by our mechanisms over ARED.

Packet reordering is generally considered to have negative impact on network performance. In this thesis, the packet reordering is used to assist TCP to recover faster in RED-enabled packet switched networks. The RED queue management prevents networks from congestion by dropping packets with a probability earlier than the time when congestion would actually occur. After a RED router drops a packet, packer reordering is introduced during TCP¡¦s recovery process. A new, simple buffer mechanism, called RED with Recovery Queue or R2Q, is proposed to create this type of packet reordering on behalf of TCP with the objective of accelerating TCP¡¦s recovery and thus improving the overall network performance. In R2Q, the original RED queue is segmented into two sub-queues. The first sub-queue remains the function of the original RED while the second picks up the packets discarded by the first. Then, scheduling of the second-chance transmission of the packets in the secondary sub-queue is the key in achieving our objective. In this thesis, we considered two scheduling schemes: priority and weighted round robin. To evaluate the performance of R2Q with these two scheduling schemes, we implemented and evaluated them in the J-Sim network simulation environment. The well-known dumbbell network topology was adopted and we varied different parameters, such as round-trip time, bottleneck bandwidth, buffer size, WRR weight and so on, in order to understand how R2Q performs under different network configurations. We found that R2Q is more effective in the networks of sufficient buffer and larger product of RTT and bandwidth. With WRR, we may achieve as much as 2% improvement over the original RED. The improvement may be more in networks of even higher speed.

There is a rapidly growing number of advanced applications running over the internet that requires ultra-low latency and high throughput. Bufferbloat is one of the most known problems which add delay in the form of packets being enqueued into large buffers before being transmitted. This has been solved with the developments of various Active Queue Management (AQM) schemes to control how large the queue buffers are allowed to grow. Another aspect that is important today is how the available bandwidth can be shared between applications with different priorities. The Per-Packet Value (PPV) concept has been presented as a solution for resource-sharing by marking packets according to predefined marking policies. The packet value will be taken into consideration to make drop/mark decisions, which leads to higher packet values being prioritized at bottleneck links.  In this thesis, a design of a packet value-based AQM on a programmable Barefoot Tofino switch will be presented. It will use a combination of the Proportional Integral Controller (PIE) AQM scheme and the PPV concept to make drop decisions when queuing delay is discovered. Packet value statistics are collected through the P4 programmable data plane to maintain knowledge about the distribution of packet values. With the dropping probability calculated through the PIE AQM scheme, a decision can be made about which packets should be dropped.  An evaluation shows that with the implemented PV AQM, a low queuing delay can be achieved by dropping an appropriate amount of packets. It also shows that the PV AQM controls the resource-sharing between different traffic flows according to a predefined marking policy.

New internet services require low and stable latency, which is difficult to provide with traditional routers and queuing mechanisms. Current routers aim to provide high throughput using large buffers causing considerable network latency under load. Recently, Active Queue Management (AQM) algorithms have been proposed to reduce such problem by actively controlling queue lengths to maintain target latencies. However, AQMs are difficult to implement in switching Application-Specific Integrated Circuits (ASIC) due to inherent architectural constraints. On the other hand, resource sharing is another important goal aiming to differentiate traffic and allocating more resources to different traffic types.  The objective of this thesis is to implement the AQM algorithm Proportional Integral Controller Enhanced (PIE) with a packet marking based resource sharing concept Per Packet Value (PPV) on a programmable switching ASIC using the novel network programmability concept P4. Our solution is designed to maintain low and controllable latency and to utilize the bottleneck link efficiently, while observing the bandwidth sharing properties of the marking scheme. Our goal is to show that Data Plane native implementations of PPV PIE using the Tofino is possible without severely limiting performance or accuracy. The solution places the computation of PIE's drop probability estimation on a timer in the Data Plane utilizing a state machine, packet mirroring, packet recirculation and approximative arithmetics implemented by lookup tables. Additionally, a small control loop is required in order to update lookup tables based on packet statistics from the Control Plane.  In our evaluation using a Tofino based testbed, we evaluate the impact of different parameters on both Control Plane latency, Data Plane throughput and delay for both static and dynamic traffic scenarios. Our results demonstrate commendable performance in terms of controlling queuing delay, effective throughput and bandwidth share when taking operator policy in regard.