Reliable transport protocols such as TCP are tuned to perform well in traditional networks where packet losses occur mostly because of congestion. However, networks with wireless and other lossy links also suffer from significant losses due to bit errors and handoffs. TCP responds to all losses by invoking congestion control and avoidance algorithms, resulting in degraded end-to-end performance in wireless and lossy systems. In this paper, we compare several schemes designed to improve the performance of TCP in such networks. We classify these schemes into three broad categories: end-to-end protocols, where loss recovery is performed by the sender; link-layer protocols, that provide local reliability; and split-connection protocols, that break the end-to-end connection into two parts at the base station. We present the results of several experiments performed in both LAN and WAN environments, using throughput and goodput as the metrics for comparison. Our results show that a reliable link-layer protocol that is TCP-aware provides very good performance. Furthermore, it is possible to achieve good performance without splitting the end-to-end connection at the base station. We also demonstrate that selective acknowledgments and explicit loss notifications result in significant performance improvements.

This paper presents a novel framework for managing network congestion from an end-to-end perspective. Our work is motivated by several trends in traffic patterns that threaten the long-term stability of the Internet. These trends include the use of multiple independent concurrent flows by Web applications and the increasing use of transport protocols and applications that do not adapt to congestion. We present an end-system architecture centered around a Congestion Manager (CM) that ensures proper congestion behavior and allows applications to easily adapt to network congestion. Our framework integrates congestion management across all applications and transport protocols. The CM maintains congestion parameters and exposes an API to enable applications to learn about network characteristics, pass information to the CM, and schedule data transmissions. Internally, it uses a stable rate-based control algorithm, a scheduler to regulate transmissions, and a lightweight loss-resilient protocol to elicit feedback from receivers. Its ratebased scheme uses additive increase/multiplicative decrease, combined with a novel exponential aging scheme when receiver feedback is infrequent, to obtain both stable network behavior and good application performance.

Abstract — Many Internet protocols and operational procedures use measurements to guide future actions. This is an effective strategy if the quantities being measured exhibit a degree of constancy: that is, in some fundamental sense, they are not changing. In this paper we explore three different notions of constancy: mathematical, operational, and predictive. Using a large measurement dataset gathered from the NIMI infrastructure, we then apply these notions to three Internet path properties: loss, delay, and throughput. Our aim is to provide guidance as to when assumptions of various forms of constancy are sound, versus when they might prove misleading. I.

The rapid growth of the World Wide Web in recent years has caused a significant shift in the composition of Internet traffic. Although past work has studied the behavior of TCP dynamics in the context of bulk-transfer applications and some studies have begun to investigate the interactions of TCP and HTTP, few have used extensive realworld traffic traces to examine the problem. This interaction is interesting because of the way in which current Web browsers use TCP connections: multiple concurrent short connections from a single host. In this paper, we analyze the way in which Web browsers use TCP connections based on extensive traffic traces obtained from a busy Web server (the official Web server of the 1996 Atlanta Olympic games). At the time of operation, this Web server was one of the busiest on the Internet, handling tens of millions of requests per day

Abstract—This paper presents a detailed analysis of traces of domain name system (DNS) and associated TCP traffic collected

In this paper, we study the dynamics of the MSNBC news site, one of the busiest Web sites in the Internet today. Unlike many other efforts that have analyzed client accesses as seen by proxies, we focus on the server end. We analyze the dynamics of both the server content and client accesses made to the server. The former considers the content creation and modification process while the latter considers page popularity and locality in client accesses. Some of our key results are: (a) files tend to change little when they are modified, (b) a small set of files tends to get modified repeatedly, (c) file popularity follows a Zipf-like distribution with a parameter ff that is much larger than reported in previous, proxy-based studies, and (d) there is significant temporal stability in file popularity but not much stability in the domains from which clients access the popular content. We discuss the implications of these findings for techniques such as Web caching (including cache consisten...

This document proposes a new TCP mechanism that can be used to more effectively recover lost segments when a connection's congestion window is small, or when a large number of segments are lost in a single transmission window. The "Limited Transmit" algorithm calls for sending a new data segment in response to each of the first two duplicate acknowledgments that arrive at the sender. Transmitting these segments increases the probability that TCP can recover from a single lost segment using the fast retransmit algorithm, rather than using a costly retransmission timeout. Limited Transmit can be used both in conjunction with, and in the absence of, the TCP selective acknowledgment (SACK) mechanism [RFC2018]. 1 Introduction A number of researchers have observed that TCP's loss recovery strategies do not work well when the congestion window at a TCP sender is small. This can happen, for instance, because there is only a limited amount of data to send, or because of the limit...

Web browsing is characterized by short and bursty data transfers interspersed by idle periods. The TCP protocol yields poor performance for such a workload because the TCP slow start procedure, which is initiated both at connection start up and upon restart after an idle period, usually requires several round trips to probe the network for bandwidth. When a transfer is short in length, this leads to poor bandwidth utilization and increased latency, which limit the performance benefits of techniques such as P-HTTP. In this paper, we present a new technique, which we call TCP fast start, to speed up short Web transfers. The basic idea is that the sender caches network parameters to avoid paying the slow start penalty for each page download. However, there is the risk of performance degradation if the cached information is stale. The two key contributions of our work are in addressing this problem. First, to shield the network as a whole from the ill-effects of stale information, packets...

There is much interest in using network measurements for both modeling and operational purposes. In this paper we focus on the fundamental question of the stationarity of such measurements. That is, to what extent are past measurements a good predictor of the future? We used the NIMI infrastructure and a set of public traceroute servers to capture large measurement datasets of three quantities: routing, packet loss, and TCP throughput. We apply statistical tests to attempt to develop sound characterizations of the stationarity of these data sets, and discuss several types of nonstationarity. 1 Introduction In recent years there has been a surge of interest in network measurements. These measurements have deepened our understanding of network behavior and led to more accurate and qualitatively different models of network traffic. Network measurements are also used operationally by various protocols to guide network usage. For instance, RLM [MJV96] and equation-based congestion control...

Abstract not found