Search

One of the new challenges associated with full-duplex Ethernet is that of providing for network congestion control. The IEEE 802.3x Standard does not specify the switch architecture for implementing PAUSE flow control or at what point a MAC Control entity actually generates a PAUSE frame. This is an implementation issue that is product specific and as a result many different switch architectures are possible. There are also a number of limitations of the PAUSE flow control mechanism when implemented in Ethernet switches. These issues have not been adequately addressed in the literature. In addition, multimedia traffic such as real-time voice and streaming video are now being deployed over switched Ethernet networks, thus calling for congestion control with service differentiation for the various classes of traffic. Here we examine current Ethernet switch architectures and show that the PAUSE flow control when implemented in these architectures does not provide service selectivity and differentiation, making it unsuitable for real-time traffic.

This article discusses the evolving transport architecture, covering some of the synchronization distribution problems to many endpoints where mobile backhaul and TDM emulation occur. It shows how synchronous Ethernet fits into both the Ethernet and synchronization architectures, and discusses how this helped development in standardization bodies. Standardization allows the key building blocks of Ethernet silicon and specific timing devices to be developed, which allows a robust system implementation to be constructed while allowing interworking with and migration from existing SONET/SDH-based transport infrastructure. Finally, results are shown that indicate a very high level of performance is achievable with Synchronous Ethernet not subject to the normal packet delay variation and traffic load conditions that can occur in packet based networks.

Most high-speed links do not have adequate buffering and as a result Active Queue Management (AQM) schemes that utilize queue size information for congestion control cannot be effectively applied on these links. A high-speed link will, typically, have small buffers in relation to the bandwidth-delay product of the link. In this paper we argue that rate-based AQM schemes be used for such links. The goal here is to match the aggregate rate of the active TCP connections to the available capacity while maintaining minimal queue size and high link utilization. The AQM scheme described here employs a Proportional-Integral (PI) control strategy and explicitly takes into account the time delay in the control process.

Due to Ethernet's ubiquity, simplicity, scalability and cost effectiveness there is significant customer demand for Ethernet-based access and transport in the metropolitan network. Many service providers have recognized this need and are currently establishing Ethernet-based services to meet this demand. The migration to all-Ethernet access will not be instantaneous since many customers currently have legacy TDM access interfaces on their routers and PBX equipment. Circuit Emulation Services (CES) over Ethernet provides TDM circuit emulation to support TDM traffic such as T1/E1, T3/E3, OC3/12, etc. This two-part paper presents the application of CES over Ethernet as well as a new technology that addresses the issues associated with clock recovery and synchronization in an Ethernet network with its inherent network jitter. Part 1 describes a clock synchronization technique where a transmitter periodically sends explicit time indications or timestamps to a receiver to enable the receiver to synchronize its local clock to the transmitter's clock.

There is current interest in differentiated service architectures where packets with different priorities can share the same queue. In the case of congestion, packets marked with higher drop probability are preferentially dropped in order to make buffer room for packets marked with lower drop probability. Active queue management (AQM) based on randomized packet dropping has become a key component of this packet forwarding model. This paper extends our previously developed AQM algorithm called DRED with multiple packet drop precedence to allow for priority treatment of traffic in a network. The main advantage of using the DRED algorithm is the lower parameter configuration complexity it offers and the ease of configuration for a wide range of network conditions.

Explicit transmission control protocol (TCP) window control through the modification of the receiver’s advertised window in ACK packets is one of the ways intermediate network elements can contribute to the end-to-end TCP control. The TCP receiver’s advertised window, which indicates the level of the receive buffer of a TCP connection, limits the maximum window and consequently the throughput that can be achieved by the TCP sender. Thus, appropriate reduction of the advertised window by intermediate network elements can control the number of packets sent from a TCP sender. This paper describes a TCP window control scheme for a shared memory system with multiple queues. A dynamic buffer threshold, computed using a simple recursive algorithm, is used to dynamically allocate buffer space to the queues.

Most active queue management schemes maintain an average of the queue length which they use together with a number of queue thresholds to detect congestion. However, the setting of the queue thresholds is problematic because the required buffer size for good sharing among TCP connections is dependent on the number of TCP connections using the buffer. This paper describes an improved active queue management scheme which dynamically changes its threshold settings as the number of connections and system load changes. This technique allows network devices to effectively control packet losses and TCP timeouts while maintaining high link utilization.

A multiple (priority) queueing system allows a network node to manage the queueing of packets in such a way that higher priority packets will always be served first, low priority packets will be discarded when the queue is full, and for same-priority packets any interference between them will be prevented. This paper describes a TCP window control scheme for a shared memory device that has buffer memory logically organized into multiple queues. To handle changing queue traffic loads, the shared memory device uses a dynamic buffer threshold mechanism to allocate buffer space to the queues. The TCP window control scheme allows the receiver's advertised window size in ACK packets to be modified at the network queue in order to maintain the queue size at a computed dynamic threshold. Copyright © 2003 John Wiley & Sons, Ltd.

In explicit TCP rate control, the receiver's advertised window size in acknowledgment (ACK) packets can be modified by intermediate network elements to reflect network congestion conditions. The TCP receiver's advertised window (i.e. the receive buffer of a TCP connection) limits the maximum window and consequently the throughput that can be achieved by the sender. Appropriate reduction of the advertised window can control the number of packets allowed to be sent from a TCP source. This paper evaluates the performance of a TCP rate control scheme in which the receiver's advertised window size in ACK packets are modified in a network node in order to match the generated load to the assigned bandwidth in the node. Using simulation and performance metrics such as the packet loss rates and the cumulative number of TCP timeouts, we examine the service improvement provided by the TCP rate control scheme to the users. The modified advertised windows computed in the network elements and the link utilization are also examined. Copyright © 2002 John Wiley & Sons, Ltd.

The transmission control protocol (TCP) receiver's advertised window (i.e. the receive buffer of a TCP connection) limits the maximum window and consequently the throughput that can be achieved by the sender. Thus, the idea behind the technique generally known as 'TCP rate control' is to match the offered network load to the available resources by modifying at an intermediate network device, the receiver's advertised window in TCP acknowledgments returning to the sources. In this paper, we propose a new TCP rate control scheme for a shared buffer where the buffer is logically organized into multiple queues. In the scheme, a dynamic buffer threshold is used to ensure efficient and fair usage of buffer memory among the queues. Conventional schemes allocate buffer space to each queue through the use of static buffer thresholds. This can result in unnecessary packet drops which lead to poor network performance since congested or heavily loaded queues cannot gain access to buffers not utilized by lightly loaded queues.

In this paper, we evaluate the performance of TCP in a network where the IEEE 802.3x flow control and the ATM Available Bit Rate flow control mechanisms are interworked. The objective is to examine whether an interworking of the two flow control mechanisms can improve the performance of upper-layer protocols like TCP which is an end-to-end connection-based transport protocol. The simulation studies show that the two interworked flow control mechanisms result in high link utilization while providing very low loss rates when used in conjunction with TCP. The flow control mechanisms complement each other.

We describe in this paper a new TCP ACK pacing scheme that dynamically tunes its behavior to account for variations or changes in the network load. The scheme does not require the knowledge of when TCP is in the slow-start or congestion avoidance phase to determine the proper ACK pacing rate.

In TCP, congestion control as well as error recovery are implemented by a sliding window. The dynamics of TCP specifically, a mismatch between the TCP window and the bandwidth-delay product of the network can sometimes cause the network switches or routers to accumulate large queues, resulting in buffer overflows, reduced throughput, unfairness and underutilization. It is generally accepted that there is a limit to how much control can be accomplished from the congestion control mechanisms in the end systems. Some mechanisms are needed in the intermediate network elements to complement the endpoint congestion avoidance mechanisms. We describe in this paper, a new TCP rate control scheme based on a simple recursive algorithm. The idea behind the algorithm is to match the offered network load to the available resources by modifying at an intermediate network element, the receiver’s advertised window in TCP acknowledgments returning to the source. We show through simulations that the scheme can efficiently control TCP traffic to limit queue buildups and buffer requirements at the network nodes, resulting in significant improvements in delay, packet loss rates, fairness in the distribution of the maximum achievable window size, and link utilization.

The basic idea behind active queue management schemes such as random early detection (RED) is to detect incipient congestion early and to convey congestion notification to the end-systems, allowing them to reduce their transmission rates before queues in the network overflow and packets are dropped. The basic RED scheme (and its newer variants) maintains an average of the queue length which it uses together with a number of queue thresholds to detect congestion. RED schemes drop incoming packets in a random probabilistic manner where the probability is a function of recent buffer fill history. The objective is to provide a more equitable distribution of packet loss, avoid the synchronization of flows, and at the same time improve the utilization of the network. The setting of the queue thresholds in RED schemes is problematic because the required buffer size for good sharing among TCP connections is dependent on the number of TCP connections using the buffer. This paper describes a technique for enhancing the effectiveness of RED schemes by dynamically changing the threshold settings as the number of connections (and system load) changes. Using this technique, routers and switches can effectively control packet losses and TCP timeouts while maintaining high link utilization.

This paper describes the Dynamic-RED (DRED) algorithm, an active queue management algorithm for TCP/IP networks. In random early detection (RED), one of the goals is to stabilize the queue lengths in routers. However, the current version of RED does not succeed in this goal because the equilibrium queue length strongly depends on the number of active TCP connections. Using a simple control-theoretic approach, DRED randomly discards packets with a load-dependent probability when a buffer in a router gets congested. Over a wide range of load levels, DRED is able to stabilize a router queue occupancy at a level independent of the number of active TCP connections. The algorithm achieves this without estimating the number of active TCP connections or flows and without collecting or analyzing state information on individual flows. The benefits of stabilized queues in a network are high resources utilization, bounded delays, more certain buffer provisioning, and traffic-load-independent network performance in terms of traffic intensity and number of connections.

To help improve the performance of congestion avoidance protocols like Transmission Control Protocol (TCP) and to limit the impact of non-adaptive UDP-based applications, the Internet Engineering Task Force has recommended the widespread deployment of active queue management schemes. We describe in this paper a new Random Early Detection (RED) algorithm for congestion control called Dynamic-RED (DRED) from a gradient optimization perspective. One of the goals of RED schemes is to stabilize the queue lengths in routers. However, the current version of RED does not succeed in this goal because the equilibrium queue length strongly depends on the number of active TCP connections. Using a simple optimization technique, DRED randomly discards packets with a load-dependent probability when a buffer in a router gets congested. DRED is also able to stabilize a router queue occupancy at a level independent of the number of active connections over a wide range of load levels. This is done without estimating the number of active TCP connections or flows and without collecting or analyzing state information on individual flows.

Enterprise networks using Ethernet switches at the edge of the network and ATM switches in the core of the network have become a common model in the industry. In such networking environment, TCP is used as the end-to-end flow control mechanism (transport layer) and ABR as the edge-to-edge flow control mechanism (link layer). It is well known that ABR pushes congestion to the edge of the network, thereby alleviating congestion in the ATM network but resulting in buffer overflows and packet losses at the interworking Ethernet/ATM unit, thus degrading end-to-end performance. We investigate the interworking behavior between the TCP flow control mechanism and ABR flow control mechanism, and show that appropriate buffer provisioning in the edge devices results in acceptable TCP performance. However, fairness between competing connections is not perfectly achieved. We also show that different TCP flavors exhibit distinctive performance behavior.

In TCP (transmission control protocol), congestion control as well as error recovery are implemented by a sliding window. The dynamics of TCP (specifically, a mismatch between the TCP window and the bandwidth-delay product of the network) can sometimes cause the network switches or routers to accumulate large queues, resulting in buffer overflows, reduced throughput, unfairness and underutilization. It is generally accepted that there is a limit as to how much control can be accomplished from the congestion control mechanisms in the end systems. Some mechanisms are thus needed in the intermediate network elements to complement the endpoint congestion avoidance mechanisms. Network layer enhancements such as scheduling mechanisms and packet drop policies have been proposed which are aimed at improving fairness and throughput of the competing endpoint applications. We describe a new TCP rate control scheme based on a simple recursive algorithm. The idea behind the algorithm is to match the network load to the available resources by modifying at an intermediate network element, the receiver's advertised window in TCP acknowledgments returning to the sources. The scheme can be implemented in a router or switch for bandwidth management and does not require knowledge of network delays or maintenance of the per-flow state.