The combination of loss-based TCP and drop-tail routers often results in full buffers, creating large queueing delays. The challenge with parameter tuning and the drastic consequence of improper tuning have discouraged network administrators from enabling AQM even when routers support it. To address this problem, we propose a novel design principle for AQM, called the pricing-link-by-time (PLT) principle. PLT increases the link price as the backlog stays above a threshold β, and resets the price once the backlog goes below β. We prove that such a system exhibits cyclic behavior that is robust against changes in network environment and protocol parameters. While β approximately controls the level of backlog, the backlog dynamics are invariant for β across a wide range of values. Therefore, β can be chosen to reduce delay without undermining system performance. We validate these analytical results using packet-level simulation.

The congestion control mechanisms in the standardized Transmission Control Protocol (TCP) may misinterpret packet reordering as congestive loss, leading to spurious congestion response and under-utilization of network capacity. Therefore, many TCP enhancements have been proposed to better differentiate between packet reordering and congestive loss, in order to enhance the reordering robustness (RR) of TCP. Since such enhancements are incrementally deployed, it is important to study the interactions of TCP flows with heterogeneous RR. This paper presents the first systematic study of such interactions by exploring how changing RR of TCP flows influences the bandwidth sharing among these flows. We define the quantified RR (QRR) of a TCP flow as the probability that packet reordering causes congestion response. We analyze the variation of bandwidth sharing as QRR changes. This leads to the discovery of several interesting properties. Most notably, we discover the counter-intuitive result that changing one flow's QRR does not affect its competing flows in certain network topologies. We further characterize the deviation, from the ideal case of bandwidth sharing, as RR changes. We find that enhancing RR of a flow may increase, rather than decrease, the deviation in some typical network scenarios.

The class of additive-increase/multiplicative-decrease (AIMD) algorithms constitutes a key mechanism for congestion control in modern communication networks, like the current Internet. The algorithmic behaviour may, however, be distorted when wireless links are present. Specifically, spurious window reductions may be triggered due to packet reordering and non-congestive loss. In this paper, we develop a framework for AIMD in TCP to analyze the aforementioned problem. The framework enables a systematic analysis of the existing AIMD-based TCP variants and assists in the design of new TCP variants. It classifies the existing AIMD-based TCP variants into two main streams, known as compensators and differentiators, and develops a generic expression that covers the rate adaptation processes of both approaches. It further identifies a new approach in enhancing the performance of TCP, known as the compensation scheme. A tax-rebate approach is proposed as an approximation of the compensation scheme, and used to enhance the AIMD-based TCP variants to offer unified solutions for effective congestion control, sequencing control, and error control. In traditional wired networks, the new family of TCP variants with the proposed enhancements automatically preserves the same inter-flow fairness and TCP friendliness. We have conducted a series of simulations to examine their performance under various network scenarios. In most scenarios, significant performance gains are attained.

In order to maintain a small, stable backlog at the router buffer, active queue management (AQM) algorithms drop packets probabilistically at the onset of congestion, leading to backoffs by Transmission Control Protocol (TCP) flows. However, wireless losses may be misinterpreted as congestive losses and induce spurious backoffs. In this paper, we raise the basic question: Can AQM maintain a stable, small backlog under wireless losses? We find that the representative AQM, random early detection (RED), fails to maintain a stable backlog under time-varying wireless losses. We find that the key to resolving the problem is to robustly track the backlog to a preset reference level, and apply the control-theoretic vehicle, internal model principle, to realize such tracking. We further devise the integral controller (IC) as an embodiment of the principle. Our simulation results show that IC is robust against time-varying wireless losses under various network scenarios.

In wireless networks, TCP performs unsatisfactorily since packet reordering and random losses may be falsely interpreted as congestive losses. This causes TCP to trigger fast retransmission and fast recovery spuriously, leading to under-utilization of available network resources. In this paper, we propose a novel TCP variant, known as TCP for non-congestive loss (TCP-NCL), to adapt TCP to wireless networks by using more reliable signals of packet loss and network overload for activating packet retransmission and congestion response, separately. TCP-NCL can thus serve as a unified solution for effective congestion control, sequencing control, and loss recovery. Different from the existing unified solutions, the modifications involved in the proposed variant are limited to sender-side TCP only, thereby facilitating possible future wide deployment. The two signals employed are the expirations of two serialized timers. A smart TCP sender model has been developed for optimizing the timer expiration periods. Our simulation studies reveal that TCP-NCL is robust against packet reordering as well as random packet loss while maintaining responsiveness against situations with purely congestive loss.

The problems of TCP packet reordering and random loss over wireless networks have motivated the development of a number of TCP variants. However, most of these variants focus on resolving only one of the aforementioned two problems. A few unified solutions for both problems generally extend beyond the scope of the transport layer. In this paper, we propose a new TCP variant, known as TCP for non-congestion loss (TCP-NCL), to tackle both problems under one compact framework. Different from previous unified solutions, the modifications are limited to sender-side TCP only, thereby facilitating possible future wide deployment. A retransmission decision timer and a congestion response decision timer have been installed to trigger packet retransmission and congestion response, respectively. Our simulation studies reveal that TCP-NCL is robust against packet reordering as well as random packet loss while maintaining responsiveness against situations with purely congestive loss.