Approval for the Report and Comprehensive Examination: Committee:

The packet losses imposed by IP networks can cause long and erratic recovery delays, since senders must often use conservative loss detection and retransmission mechanisms. This paper proposes a model to explain and predict loss rates for TCP traffic. Based on that model, the paper describes a new router buffering algorithm, Flow-Proportional Queuing (FPQ), that handles heavy TCP loads without imposing high loss rates. FPQ controls TCP by varying the router's queue length in proportion to the number of active TCP connections. Simulation results show that FPQ produces the same average transfer delays as existing schemes, but makes the delays more predictable and fairer.

The vision of ad-hoc networking with Bluetooth includes the concept of devices participating in multiple "piconets" and thereby forming a "scatternet". However, the details of scatternet support for Bluetooth are not specified yet. This paper presents a scheme for Bluetooth scatternet operation that adapts to varying traffic patterns. Basing on sniff mode, it does not require substantial modification of the current Bluetooth specification and may thus be incorporated into currently available Bluetooth products. We present simulation results that confirm the applicability of our approach to realistic scenarios.

The abstract system design and a prototype implementation of an admission control system is described. The system exploits the load signal generated by low complexity active queue management schemes at internal nodes to carry out flow admission control at edge gateways. The location of functional components is studied and appropriate signalling extensions to exchange load information are presented. The suitability of such signalling extensions for admission control and traffic regulation are discussed in the context of RSVP signalling. A number of questions of detail which are usually ignored by existing theory and simulation work, are examined and solutions are presented. Certain modifications to traffic control algorithms at both internal packet marking nodes and edge gateways are proposed and discussed. The functionality and correct system operation are demonstrated by experiments using the software prototype. Further, a variety of marking algorithms is compared experimentally to assess the suitability of their respective load signal for admission control of inelastic traffic and load-adaptive traffic regulation.

TCP Westwood (TCPW) is a sender-side only modification of TCP Reno congestion control, which exploits end-to-end bandwidth estimation to properly set the values of slow-start threshold and congestion window after a congestion episode. This paper aims at showing via both mathematical modeling and extensive simulations that TCPW significantly improves fair sharing of high-speed networks capacity and that TCPW is friendly to TCP Reno. Moreover, we propose EASY RED, which is a simple Active Queue management (AQM) scheme that improves fair sharing of network capacity especially over high-speed networks. Simulation results show that TCP Westwood provides a remarkable Jain's fairness index increment up to 200% with respect to TCP Reno and confirm that TCPW is friendly to TCP Reno. Finally, simulations show that Easy RED improves fairness of Reno connections more than RED, whereas the improvement in the case of Westwood connections is much smaller since Westwood already exhibits a fairer behavior by itself. 1.

Packet networks require queues (buffers) to absorb short term arrival rate fluctuations. Yet network implementors have always observed that queues at bottlenecks tend to fill and stay filled, which contributes extra delay and removes the ability to absorb bursts. In [1] Floyd and Jacobson proposed the RED (Random Early Detection) active queue management algorithm. RED is simple, robust and quite effective at reducing persistent queues. However, while it has been used widely and successfully on Internet routers, [1] offers little guidance on how to set configuration parameters and RED has gained the reputation of being very difficult to tune. This paper develops RED in different way, treating it as a servo control loop and deriving all the loop parameters from measurable properties of a router. The result is a `self-tuning' RED whose parameters are completely determined by the queue output bandwidth (average departure rate). This new RED performs substantially better than the original version and works for a much wider variety of traffic and link bandwidths. It also admits a substantially simpler and more efficient implementation, one particularly well suited for ASIC forwarding engines. Please note: This is an early draft of an in-progress paper. Several important sections are still missing and the simulation data needs to be reorganized so that the story it tells is clearer. 1.0

Routers making use of Random Early Detection (RED) queueing take action to notify sources of growing congestion levels in the network before their resources are exhausted. The RED system hinges on two calculations: tracking the average queue size and the probability that an incoming packet is marked for congestion. These two calculations can be done in terms of the number of packets arriving at the router or in terms of the size of those packets (in bytes). Intuitively, these calculation methods offer different costs and benefits to traffic. This paper quantitatively assesses the impact of using the different queueing and marking methods on the performance of traffic traversing a RED gateway. We show that in some cases the calculation method makes a difference in the performance of the system, while in other cases the choice has little impact. We also provide a framework for rating the RED variants in particular situations in an attempt to aid in the choice of variant to use in a specific situations.

As flows of traffic traverse a network, they share with other flows a variety of resources such as links, buffers and router CPUs in their path. Fairness is an intuitively desirable property in the allocation of resources in a network shared among flows of traffic from different users. While fairness in bandwidth allocation over a shared link has been extensively studied, overall end-to-end fairness in the use of all the resources in the network is ultimately the desired goal. End-to-end fairness becomes especially critical when fair allocation algorithms are used as a component of the mechanisms used to provide end-to-end quality-of-service guarantees. This paper seeks to answer the question of what is fair when a set of traffic flows share multiple resources in the network with a shared order of preference for the opportunity to use these resources. We present the Generalized Principle of Fairness (GPF), a powerful extension of any of the classic notions of fairness such as max-min fairness, proportional fairness and utility max-min fairness defined over a single resource. We illustrate this principle by applying it to a system model with a buffer and an output link shared among competing flows of traffic. To complete our illustration of the applicability of GPF, we propose a measure of fairness and evaluate representative buffer allocation algorithms based on this measure. Besides buffer allocation, GPF may also be used in other contexts in data communication networks and operating system design.

Abstract not found

Abstract—The need for service resilience is leading to a steadily growing number of multi-homed Internet sites. In consequence, this results in a growing demand for utilising multiple Internet accesses simultaneously, in order to improve application payload throughput during normal operation. Multi-path Transport Layer protocol extensions – like Multi-Path TCP (MPTCP) for TCP and Concurrent Multipath Transfer for SCTP (CMT-SCTP) – allow applications to make use of such network topologies. However, since TCP – which constitutes the basis of most Internet applications – and its congestion control procedures have been designed under the assumption of single-homed sites, fairness issues may arise by the usage of multipath transport. These issues are addressed by advanced congestion control approaches, which have already been examined by simulations. However, real-life network measurements are missing. In this paper, we perform an experimental proof-of-concept evaluation of several multipath congestion control strategies, which are currently under discussion within the IETF in the context of MPTCP as well as CMT-SCTP. Particularly, we validate effects that have been observed in simulations, in order to trigger further discussions on multipath congestion control. Also, our goal is to provide insight into the different approaches to support the ongoing IETF standardisation activities on multipath transport protocols. 1234