In this dissertation, we present a novel service architecture for the Internet, which reconciles application demand for strong service guarantees with the need for low computational overhead in network routers. The main contribution of this dissertation is the definition and realization of a new service, called Quantitative Assured Forwarding, which can offer absolute and relative differentiation of loss, service rates, and packet delays to classes of traffic. We devise and analyze mechanisms that implement the proposed service, and demonstrate the effectiveness of the approach through analysis, simulation and measurement experiments in a testbed network. To enable

Abstract — This paper examines congestion control issues for TCP flows that require in-network processing on the fly in network elements such as gateways, proxies, firewalls and even routers. Applications of these flows are increasingly abundant in the future as the Internet evolves. Since these flows require use of CPUs in network elements, both bandwidth and CPU resources can be a bottleneck and thus congestion control must deal with “congestion ” on both of these resources. In this paper, we show that conventional TCP/AQM schemes can significantly lose throughput and suffer harmful unfairness in this environment, particularly when CPU cycles become more scarce (which is likely the trend given the recent explosive growth rate of bandwidth). As a solution to this problem, we establish a notion of dual-resource proportional fairness and propose an AQM scheme, called Dual-Resource Queue (DRQ), that can closely approximate proportional fairness for TCP Reno sources with in-network processing requirements. DRQ is scalable because it does not maintain perflow states while minimizing communication among different resource queues, and is also incrementally deployable because of no required change in TCP stacks. The simulation study shows that DRQ approximates proportional fairness without much implementation cost and even an incremental deployment of DRQ at the edge of the Internet improves the fairness and throughput of these TCP flows. Our work is at its early stage and might lead to an interesting development in congestion control research. I.

This paper introduces and studies a new queue management framework via the use of connection level information already embedded in existing data traffic. Our goal is to improve the system performance and resource utilization at times of intense network congestion. Under this framework, data packets and connection establishment packets are queued separately. Different queueing and dropping policies can then be applied to them, thus the term dual-queue management. This framework is stateless and does not require per-flow queueing or flow counting. Using examples of such policies our initial simulation shows that this scheme can lead to much higher performance and network resource utilization during severe congestion. It also leads to more robust and scalable network design. Moreover, it provides an attractive mechanism for network provisioning via parameter tuning.

Recently, several gateway-based congestion control mechanisms have been proposed to support the endto -end congestion control mechanism of TCP (Transmission Control Protocol). In this paper, we focus on RED (Random Early Detection), which is a promising gateway-based congestion control mechanism. RED randomly drops an arriving packet with a probability proportional to its average queue length (i.e., the number of packets in the buffer). However, it is still unclear whether the packet marking function of RED is optimal or not. In this paper, we investigate what type of packet marking function, which determines the packet marking probability from the average queue length, is suitable from the viewpoint of both steady state and transient state performances. Presenting several numerical examples, we investigate the advantages and disadvantages of three packet marking functions: linear, concave, and convex. We show that, although the average queue length in the steady state becomes larger, use of a concave function improves the transient behavior of RED and also realizes robustness against network status changes such as variation in the number of active TCP connections. Moreover, by applying our analytic approach to Adaptive RED, we discuss its steady state performance and transient state performance. Our analytic results show that Adaptive RED is effective for improving steady state performance compared with the original RED, but it has little performance gain in terms of transient state performance and robustness.

Abstract. In this paper we use fixed point methods to model the behavior of a population of TCP flows traversing a network of routers implementing either Drop Tail or the RED queue management policies. We formulate a non-linear problem with the router average queue lengths as unknowns. Once the average queue lengths are obtained, other metrics such as router loss probability, TCP flow throughput, TCP flow endto-end loss rates, and average round trip time can be easily obtained. Comparison with simulation for a variety of scenarios shows that the model is quite accurate in its predictions. 1

Abstract — One of the main challenges in the internet backbone is to provide an adequate quality of service (QoS) which can be done by either adopting a suitable buffer management method or by QoS routing. One of the popular buffer management algorithms used is random early detection (RED) which falls into the category of Active Queue Management (AQM) algorithms. There are many variant algorithms based on RED including adaptive RED. In this paper two new AQM algorithms which depend on changing minimum threshold set the dropping probability manually at the maximum threshold. In the first algorithm, this changing depends on the value of the average queue length and the dropping probability at the average queue length. In the second algorithm, this change depends on the value of the dropping probability at target mean queue length. It has been shown by simulation to have a similar throughput and lower delay than either RED or adaptive RED.

Abstract — Recent studies have shown that a non-negligible number of packets face multiple congested links on Internet paths. We investigate the impact of using multiple Active Queue Management (AQM) routers on paths that consist of multiple congested links. We present the results of an ns-2 evaluation study of various AQM techniques that, in contrast to previous studies of AQM, uses a complex network topology including up to three congested links, reverse path traffic, and realistic round-trip times. We consider the effects of multiple AQM routers coupled with Explicit Congestion Notification (ECN) on the throughput of long-lived flows and response times of web traffic. We find that, especially for web traffic, performance is improved as the number of AQM routers is increased, but that significant improvements occur with even a single AQM router. I.

Several Active Queue Management (AQM) techniques for routers in the Internet have been proposed and studied during the past few years. One of the widely studied proposals, Random Early Detection (RED), involves dropping an incoming packet with some probability based on the estimated average queue length at the router. The analytical approaches to obtaining average drop probabilities in a RED enabled queue have been either based on using the instantaneous queue size for calculating the drop probability or have considered averaging with a fluid approximation. In this paper, we use a singular perturbation based approach to analyse a RED enabled queue with drop probabilities based on the estimated average queue size as has been proposed in the standard RED. The singular perturbation approach is motivated by the fact that the instantaneous and the estimated average queue lengths evolve at two different time scales. We present an analytical method to calculate the average queue size and the average drop probability for the non responsive flows. We also provide analytical expressions for the Poisson arrivals and exponential service times case.

In this paper, we propose an analytic approach of modeling a closed-loop network with multiple feedback loops using fluid-flow approximation. Specifically, we model building blocks of a network (i.e., the congestion control mechanism of TCP, propagation delay of a transmission link, and the buffer of a router) as independent continuous-time systems. By interconnecting these systems, we obtain the model for a complex closed-loop network. We improve the accuracy of analytic models for TCP congestion control and RED router by extending existing fluid-flow models. First, we obtain a block diagram for each continuous-time system using a standard CAD tool widely used in control engineering. Second, we evaluate the performance of a closed-loop network with multiple feedback loops by connecting these block diagrams. We also validate the effectiveness of our analytic approach by comparing our analytic results with simulation results. Unlike other fluid-based modeling approaches, our analytic approach is scalable and accurate; our analytic approach is scalable in terms of the number of TCP connections and routers since both input/output of all continuous-time systems are uniformly defined as a packet transmission rate. Our analytic approach is accurate since the timeout mechanism of TCP and the packet dropping algorithm of RED router are rigorously modeled in our continuous-time systems.