All Internet routers contain buffers to hold packets during times of congestion. Today, the size of the buffers is determined by the dynamics of TCP's congestion control algorithm. In particular, the goal is to make sure that when a link is congested, it is busy 100% of the time; which is equivalent to making sure its buffer never goes empty. A widely used rule-of-thumb states that each link needs a buffer of size B = RTT C, where RTT is the average round-trip time of a flow passing across the link, and C is the data rate of the link. For example, a 10Gb/s router linecard needs approximately 250ms 10Gb/s = 2.5Gbits of buffers; and the amount of buffering grows linearly with the line-rate. Such large buffers are challenging for router manufacturers, who must use large, slow, off-chip DRAMs. And queueing delays can be long, have high variance, and may destabilize the congestion control algorithms. In this paper we argue that the rule-of-thumb (B = RTT is now outdated and incorrect for backbone routers. This is because of the large number of flows (TCP connections) multiplexed together on a single backbone link. Using theory, simulation and experiments on a network of real routers, we show that a link with n flows requires no more than B = (RTT # n, for long-lived or short-lived TCP flows. The consequences on router design are enormous: A 2.5Gb/s link carrying 10,000 flows could reduce its buffers by 99% with negligible difference in throughput; and a 10Gb/s link carrying 50,000 flows requires only 10Mbits of buffering, which can easily be implemented using fast, on-chip SRAM.

In order to stem the increasing packet loss rates caused by an exponential increase in network traffic, the IETF is considering the deployment of active queue management techniques such as RED [13]. While active queue management can potentially reduce packet loss rates in the Internet, this paper shows that current techniques are ineffective in preventing high loss rates. The inherent problem with these queue management algorithms is that they all use queue lengths as the indicator of the severity of congestion.

The congestion control mechanisms used in TCP have been the focus of numerous studies and have undergone a number of enhancements. However, even with these enhancements, TCP connections still experience alarmingly high loss rates, especially during times of congestion. The IETF has addressed this problem by advocating the deployment of active queue management mechanisms, such as RED, in the network. While RED can potentially improve packet loss rates, we show that its effectiveness is highly dependent upon its operating parameters. In fact, in cases where these parameters do not match the requirements of the network load, the performance of the RED gateway can approach that of a traditional drop-tail gateway. To alleviate this problem, we propose and experiment with a self-configuring active queue management mechanism which can significantly reduce loss rates across congested links. When used in the network, this mechanism can effectively reduce packet loss while maintaining high link utilizations under the most difficult scenarios. Keywords: Congestion control, Internet, TCP, RED, queue management 1

This memo presents two recommendations to the Internet community concerning measures to improve and preserve Internet performance. It presents a strong recommendation for testing, standardization, and widespread deployment of active queue management in routers, to improve the performance of today's Internet. It also urges a concerted effort of research, measurement, and ultimate deployment of router mechanisms to protect the Internet from flows that are not sufficiently responsive to congestion notification. 1 INTRODUCTION The Internet protocol architecture is based on a connectionless end-to-end packet service using the IP protocol. The advantages of its connectionless design, flexibility and robustness, have been amply demonstrated. However, these advantages are not without cost: careful design is required to provide good service under heavy load. In fact, lack of attention to the dynamics of packet forwarding can result in severe service degradation or "Internet meltdown". T...

This paper explores the claims that TCP Vegas [2] both uses network bandwidth more efficiently and achieves higher network throughput than TCP Reno [6]. It explores how link bandwidth, network buffer capacity, TCP receiver acknowledgment algorithm, and degree of network congestion affect the relative performance of Vegas and Reno. 1 Introduction Jacobson released his TCP slow-start flow control algorithm [5] in the Tahoe distribution of bsd unix and revised it two years later for the Reno distribution [6]. Since then, researchers have implemented the RFC 1323 extensions -- bigger TCP windows and time-stamped based rtt exchange -- to improve TCP performance over high bandwidth connections. The RFC 1323 extensions do not, however, implement congestion avoidance. Last year, Brakmo, O'Malley and Peterson [2] claimed that their sender-side congestion avoidance algorithm, dubbed TCP Vegas, yielded 40-70% better throughput while retransmitting 2-5 times fewer segments than TCP Reno, both in...

Blue(SFB), a novel technique for enforcing fairness among a large number of flows. SFB scalably detects and rate-limits non-responsive flows through the use of a marking probability derived from the BLUE queue management algorithm and a Bloom filter. Using analysis and simulation, SFB is shown to effectively handle non-responsive flows using an extremely small amount of state information.

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.

A large and rapidly growing proportion of users connect to the Internet via residential broadband networks such as Digital Subscriber Lines (DSL) and cable. Residential networks are often the bottleneck in the last mile of today’s Internet. Their characteristics critically affect Internet applications, including voice-over-IP, online games, and peer-to-peer content sharing/delivery systems. However, to date, few studies have investigated commercial broadband deployments, and rigorous measurement data that characterize these networks at scale are lacking. In this paper, we present the first large-scale measurement study of major cable and DSL providers in North America and Europe. We describe and evaluate the measurement tools we developed for this purpose. Our study characterizes several properties of broadband networks, including link capacities, packet round-trip times and jitter, packet loss rates, queue lengths, and queue drop policies. Our analysis reveals important ways in which residential networks differ from how the Internet is conventionally thought to operate. We also discuss the implications of our findings for many emerging protocols and systems, including delay-based congestion control (e.g., PCP) and network coordinate systems (e.g., Vivaldi).

The space communication environment and mobile and wireless communication environments show many similarities when observed from the perspective of a transport protocol. Both types of environments exhibit loss caused by data corruption and link outage, in addition to congestion-related loss. The constraints imposed by the two environments are also similar --- power, weight, and physical volume of equipment are scarce resources. Finally, it is not uncommon for communication channel data rates to be severely limited and highly asymmetric. We are working on solutions to these types of problems for space communication environments, and we believe that these solutions may be applicable to the mobile and wireless community. As part of our work, we have defined and implemented the Space Communications Protocol Standards-Transport Protocol (SCPSTP) , a set of extensions to TCP that address the problems that we have identified. The results of our performance tests, both in the laboratory and on actual satellites, indicate that the SCPS-TP extensions yield significant improvements in throughput over unmodified TCP on error-prone links. Additionally, the SCPS modifications significantly improve performance over links with highly asymmetric data rates.

Shortest path routing protocols may suffer from congestion due to the use of a single shortest path between a source and a destination. The goal of our work is to first understand how links become overloaded in an IP backbone, and then to explore if the routing protocol, -- either in its existing form, or in some enhanced form could be made to respond immediately to overload and reduce the likelihood of its occurrenc Our method is to use extensive measurements of Sprint backbone network, measuring 138 links between September 2000 and June 2001. We find that since the backbone is designed to be overprovisioned, link overload is rare, and when it occurs, 80% of the time it is caused due to link failures. Furthermore, we find that when a link is overloaded, few (if any) other links in the network are also overloaded This suggests that deflecting packets to less utilized alternate paths could be an effective method for tackling overload. We analytically derive the condition that a network, which has multiple equal length shortest paths between every pair of nodes (as is common in the highly meshed backbone networks) can provide for loop-free deflection paths if all the link weights are within a ratio 1 + 1/(d-1)of each other; where d is the diameter of the network. Based on our measurements, the nature of the backbone topology and the careful use of link weights, we propose a deflection routing algorithm to tackle link overload where each node makes local decisions. Simulations suggest that this can be a simple and efficient way to overcome link overload without requiring any changes to the routing protocol. Link Failure, Traffic Engineering.