We describe FAST TCP, a new TCP congestion control algorithm for high-speed long-latency networks, from design to implementation. We highlight the approach taken by FAST TCP to address the four difficulties, at both packet and flow levels, which the current TCP implementation has at large windows. We describe the architecture and characterize the equilibrium and stability properties of FAST TCP. We present experimental results comparing our first Linux prototype with TCP Reno, HSTCP, and STCP in terms of throughput, fairness, stability, and responsiveness. FAST TCP aims to rapidly stabilize high-speed long-latency networks into steady, efficient and fair operating points, in dynamic sharing environments, and the preliminary results are promising.

We propose a duality model of congestion control and apply it to understand the equilibrium properties of TCP and active queue management schemes. Congestion control is the interaction of source rates with certain congestion measures at network links. The basic idea is to regard source rates as primal variables and congestion measures as dual variables, and congestion control as a distributed primal-dual algorithm carried out over the Internet to maximize aggregate utility subject to capacity constraints. The primal iteration is carried out by TCP algorithms such as Reno or Vegas, and the dual iteration is carried out by queue management such as DropTail, RED or REM. We present these algorithms and their generalizations, derive their utility functions, and study their interaction.

REM is an active queue management scheme that measures congestion not by a performance measure such as loss or delay, but by a quantity we call price. Price is computed by each link distributively using local information and is fed back to the sources through packet dropping or marking. This decoupling of congestion and performance measures allows REM to achieve high utilization with negligible delays and buffer overflow regardless of the number of. sources. We prove that REM is asymptotically stable and compare its performance with RED using simulations.

In this paper, we study cross-layer design for congestion control in multihop wireless networks. In previous work, we have developed an optimal cross-layer congestion control scheme that jointly computes both the rate allocation and the stabilizing schedule that controls the resources at the underlying layers. However, the scheduling component in this optimal crosslayer congestion control scheme has to solve a complex global optimization problem at each time, and is hence too computationally expensive for online implementation. In this paper, we study how the performance of cross-layer congestion control will be impacted if the network can only use an imperfect (and potentially distributed) scheduling component that is easier to implement. We study both the case when the number of users in the system is fixed and the case with dynamic arrivals and departures of the users, and we establish performance bounds of cross-layer congestion control with imperfect scheduling. Compared with a layered approach that does not design congestion control and scheduling together, our cross-layer approach has provably better performance bounds, and substantially outperforms the layered approach. The insights drawn from our analyses also enable us to design a fully distributed cross-layer congestion control and scheduling algorithm for a restrictive interference model.

In this paper we study the statistics of the realized throughput of elastic document transfers, accounting for the way network bandwidth is shared dynamically between the randomly varying number of concurrent flows. We first discuss the way TCP realizes statistical bandwidth sharing, illustrating essential properties by means of packet level simulations. Mathematical flow level models based on the theory of stochastic networks are then proposed to explain the observed behavior. A notable result is that first order performance (e.g., mean throughput) is insensitive with respect both to the flow size distribution and the flow arrival process, as long as “sessions ” arrive according to a Poisson process. Perceived performance is shown to depend most significantly on whether demand at flow level is less than or greater than available capacity. The models provide a key to understanding the effectiveness of techniques for congestion management and service differentiation. 1.

We discuss the relevance of fairness as a design objective for congestion control mechanisms in the Internet. Specifically, we consider a backbone network shared by a dynamic number of short-lived flows, and study the impact of bandwidth sharing on network performance. In particular, we prove that for a broad class of fair bandwidth allocations, the total number of ows in progress remains finite if the load of every link is less than one. We also show that provided the bandwidth allocation is "sufficiently" fair, performance is optimal in the sense that the throughput of the ows is mainly determined by their access rate. Neither property is guaranteed with unfair bandwidth allocations, when priority is given to one class of ow with respect to another. This suggests current proposals for a differentiated services Internet may lead to suboptimal utilization of network resources.

This article reviews the current transmission control protocol (TCP) congestion control protocols and overviews recent advances that have brought analytical tools to this problem. We describe an optimization-based framework that provides an interpretation of various flow control mechanisms, in particular, the utility being optimized by the protocol’s equilibrium structure. We also look at the dynamics of TCP and employ linear models to exhibit stability limitations in the predominant TCP versions, despite certain built-in compensations for delay. Finally, we present a new protocol that overcomes these limitations and provides stability in a way that is scalable to arbitrary networks, link capacities, and delays.

Abstract not found

We address the question of how well end-to-end transport connections perform in a satellite environment composed of one or more satellites in geostationary orbit (GEO) or low-altitude earth orbit (LEO), in which the connection may traverse a portion of the wired Internet. We first summarize the various ways in which latency and asymmetry can impair the performance of the Internet's Transmission Control Protocol (TCP), and discuss extensions to standard TCP that alleviate some of these performance problems. Through analysis, simulation, and experiments, we quantify the performance of state-of-the-art TCP implementations in a satellite environment. A key part of the experimental method is the use of traffic models empirically derived from Internet traffic traces. We identify those TCP implementations that can be expected to perform reasonably well, and those that can suffer serious performance degradation. An important result is that, even with the best satellite-optimized TCP implementations, moderate levels of congestion in the wide-area Internet can seriously degrade

The vision is tantalizing: a high-performance, scalable, and widely deployed wireless Internet that facilitates services ranging from radically new and unforeseen applications to true wireless "broadband" to residences and public spaces at rates of 10s of Mb/sec. However, while high-speed wireless access is easy to achieve in an enterprise network via low-cost IEEE 802.11 (WiFi) access points, wireless technology in public spaces is in its infancy. "Hot spots" provide high-speed wireless access, but do so in very few isolated "islands" at immense costs. Likewise, while fixed wireless (e.g. LMDS) and 3G can provide ubiquitous coverage and 3G can support mobility, throughputs can often be two orders of magnitude slower than WiFi.