Vegas is a new implementation of TCP that achieves between 40 and 70 % better throughput, with one-fifth to onehalf the losses, as compared to the implementation of TCP in the Reno distributionof BSD Unix. This paper motivates and describes the three key techniques employed by Vegas, and presents the results of a comprehensive experimental performance study—using both simulations and measurements on the Internet—of the Vegas and Reno implementations of TCP. 1

Vegas is an implementation of TCP that achieves between 37 and 71 % better throughput on the Internet, with one-fifth to one-half the losses, as compared to the implementation of TCP in the Reno distribution of BSD Unix. This paper motivates and describes the three key techniques employed by Vegas, and presents the results of a comprehensive experimental performance study—using both simulations and measurements on the Internet—of the Vegas and Reno implementations of TCP.

Network protocols are usually tested in operational networks or in simulated environments. With the former approach it is not easy to set and control the various operational parameters such as bandwidth, delays, queue sizes. Simulators are easier to control, but they are often only an approximate model of the desired setting, especially for what regards the various traffic generators (both producers and consumers) and their interaction with the protocol itself. In this paper we show how a simple, yet flexible and accurate network simulator -- dummynet -- can be built with minimal modifications to an existing protocol stack, allowing experiments to be run on a standalone system. dummynet works by intercepting communications of the protocol layer under test and simulating the effects of finite queues, bandwidth limitations and communication delays. It runs in a fully operational system, hence allowing the use of real traffic generators and protocol implementations, while solving the prob...

Based on experiments conducted in a network simulator and over real networks, this paper proposes changes to the congestion control scheme in current TCP implementations to improve its behavior during the start-up period of a TCP connection. The scheme, which includes Slow-start, Fast Retransmit, and Fast Recovery algorithms, uses acknowledgments from a receiver to dynamically calculate reasonable operating values for a sender's TCP parameters governing when and how much a sender can pump into the network. During the startup period, because a TCP sender starts with default parameters, it often ends up sending too many packets and too fast, leading to multiple losses of packets from the same window. This paper shows that recovery from losses during this start-up period is often unnecessarily time-consuming. In particular, using the current Fast Retransmit algorithm, when multiple packets in the same window are lost, only one of the packet losses may be recovered by each Fast Retransmi...

Graphs are commonly used to model the topological structure of internetworks, to study problems ranging from routing to resource reservation. A variety of graphs are found in the literature, including fixed topologies such as rings or stars, "well-known" topologies such as the ARPAnet, and randomly generated topologies. While many researchers rely upon graphs for analytic and simulation studies, there has been little analysis of the implications of using a particular model, or how the graph generation method may a ect the results of such studies. Further, the selection of one generation method over another is often arbitrary, since the differences and similarities between methods are not well understood. This paper considers the problem of generating and selecting graph models that reflect the properties of real internetworks. We review generation methods in common use, and also propose several new methods. We consider a set of metrics that characterize the graphs produced by a method, and we quantify similarities and differences amongst several generation methods with respect to these metrics. We also consider the effect of the graph model in the context of a speciffic problem, namely multicast routing.

Unlike wired networks, packets transmitted on wireless channels are often subject to burst errors which cause back to back packet losses. Most wireless LAN link layer protocols recover from packet losses by retransmitting lost segments. When the wireless channel is in a burst error state, most retransmission attempts fail, thereby causing poor utilization of the wireless channel. Furthermore, in the event of multiple sessions sharing a wireless link, FIFO packet scheduling can cause the HOL blocking effect, resulting in unfair sharing of the bandwidth. This observation leads to a new class of packet dispatching methods which explicitly take wireless channel characteristics into consideration in making packet dispatching decisions. We compare a variety of channel state dependent packet (CSDP) scheduling methods with a view towards enhancing the performance of transport layer sessions. Our results indicate that by employing a CSDP scheduler at the wireless LAN device driver level, signif...

this paper we study the effect of burst packet errors and error recovery mechanisms employed in wireless MAC protocols on the performance of transport protocols such as TCP. Most wireless LAN link layer protocols recover from packet losses by retransmitting lost segments. When the wireless channel is in a burst error state, most retransmission attempts fail, thereby causing poor utilization of the wireless channel. Furthermore, in the event of multiple sessions sharing a wireless link, FIFO packet scheduling can cause the HOL blocking effect, resulting in unfair sharing of the bandwidth. This observation leads to a new class of packet dispatching methods which explicitly take wireless channel characteristics into consideration in making packet dispatching decisions. We compare a variety of channel state dependent packet (CSDP) scheduling methods with a view towards enhancing the performance of transport layer sessions. Our results indicate that by employing a CSDP scheduler at the wireless LAN device driver level, significant improvement in channel utilization can be achieved in typical wireless LAN configurations.

Understanding the performance of the Internet’s Transmission Control Protocol (TCP) is important because it is the dominant protocol used in the Internet today. Various testing methods exist to evaluate TCP performance, however all have pitfalls that need to be understood prior to obtaining useful results. Simulating TCP is difficult because of the wide range of variables, environments, and implementations available. Testing TCP modifications in the global Internet may not be the answer either: testing new protocols on real networks endangers other people’s traffic and, if not done correctly, may also yield inaccurate or misleading results. In order for TCP research to be independently evaluated in the Internet research community there is a set of questions that researchers should try to answer. This paper attempts to list some of those questions and make recommendations as to how TCP testing can be structured to provide useful answers. 1

Simulation is a critical tool in developing, testing, and evaluating network protocols and architectures. This paper describes x-Sim, a network simulator based on the x-kernel, that is able to fully simulate the topologies and traffic patterns of large scale networks. It also gives case studies to help illustrate the capabilities and usefulness of the simulator. Finally, we consider a set of basic simulation principles in the context of running network simulations, giving a set of concrete guidelines and, most importantly, providing examples to help quantify the value of these principles as well as the cost of ignoring them. 1 Introduction We are on the brink of a new era in networking. As the Internet becomes an integral part of our daily life, mostly as a result of the large scale adoption of a new generation of distributed tools like the World-Wide-Web, major changes in the infrastructure of the Internet are being proposed. These changes are designed both to support the increasing...

In this paper we presentacouple of tools developed by theauthor on FreeBSD, and available from the author's Web page in source format. The rst one, called dummynet, is a tool designed for the performance evaluation of network protocols and applications. Despite its original design goal, there has been a lot of interest on using dummynet as a bandwidth manager in network servers. dummynet simulates the e ect of nite queues, bandwidth limitations, and queueing delays, and is embedded in the protocol stack of the host, allowing even complex experiments to be run on a single machine, using existing applications and protocol implementations. The second tool is a software implementation of an erasure code especially suited for use in network protocols. Erasure codes are used in Forward Error Correction (FEC) techniques to reduce or remove the need for retransmissions in presence of communication errors. FEC has been rarely used in network protocols, because of the encoding/decoding overhead, and also because the underlying theory of error correcting codes is generally not well known to network researchers. In this paper we discuss the theory behind a simple erasure code, and provide performance data to show that the encoding/decoding overhead is acceptable for many applications even on low-end machines. 1