Queue Management Mechanism Research Articles

An adaptive AQM to provide delay guarantees for differentiated services

RIO(RED with IN and OUT) is the primary queue management mechanism proposed for assured forwarding in the DiffServ (Differentiated Service) framework. Although RIO can generally provide bandwidth guarantees, its queuing delay is sensitive to the traffic load. This paper presents a qualitative explanation for its origin. As a solution, an Adaptive RIO for Delay (ARIO-D) is proposed to provide guaranteed delay for multimedia traffic. Simulation results show that by trading loss for delay, ARIO-D can effectively improve the robustness of RIO under different and dynamic traffic, and provide stable and differentiated performance of queuing delay without any degradation in performance of throughput.

A stable queue-based adaptive controller for improving AQM performance

Active queue management (AQM) mechanisms are designed to provide better support for end-to-end congestion control mechanisms of transmission control protocol (TCP) in TCP/IP networks. This paper introduces a stable queue-based adaptive proportional–integral (Q-SAPI) controller for AQM and presents an implementation. The starting points of our approach are the recently developed fluid-flow modeling and control theoretic interpretation of the TCP/AQM dynamics, and the recently developed fixed-gain proportional–integral (PI) controller for AQM. Q-SAPI aims to improve the transient performance of the fixed-gain PI controller while maintaining its steady-state performance over a wide range of uncertainties in round-trip time (RTT) and the number of active TCP flows. The robustness of Q-SAPI is studied in detail, which provides guidelines for selecting control parameters. Through extensive simulations, we demonstrate the ability of Q-SAPI in controlling queue length in both transient and steady states. Q-SAPI achieves this by adapting the controller gains according to the queue length.

Relative loss rate differentiation: performance of short-lived TCP flows

The relative differentiated service model provides assurances for the relative quality ordering between service classes, rather than for the actual service level in each class. In this paper, we describe a relative loss rate differentiation scheme where packet drop probabilities are determined according to an active queue management (AQM) mechanism based on random early detection (RED) in a first-in first-out (FIFO) queue, are weighted in inverse proportion to the price that the network operator assigns to each service class. Basically, we describe a scheme where relative loss rate differentiation is incorporated directly into AQM. Most TCP flows today, particularly Web flows, can be characterized as short-lived flows. Using simulations with short-lived TCP flows, we show that the scheme is very effective in ensuring relative loss rate differentiation between service classes during times of network congestion. Copyright © 2004 John Wiley & Sons, Ltd.

Subsidized RED: an active queue management mechanism for short-lived flows

In heavily congested networks, RED using parameters that provide the best link utilization produces poor response times. In this work, we devise and study the performance of a simple active queue management mechanism called Subsidized RED (SubRED) targeted at short-lived flows or fragile flows (e.g. most HTTP flows) to keep link utilization high while reducing the average flow response time. Our simulation studies have compared the performance of RED and SubRED for a network with HTTP traffic at offered loads less, close to, and more than the link capacity. Simulations show that a subsidy given to a flow provides significantly better performance in terms of HTTP request-reply delays without sacrificing the link utilization. The new scheme is very simple to implement. Combined with our cost-effective flow table based implementation, SubRED does not impose an excessive resource overhead.

PAQM: an adaptive and proactive queue management for end‐to‐end TCP congestion control

AbstractTwo functions, the congestion indicator (i.e. how to detect congestion) and the congestion control function (i.e. how to avoid and control congestion), are used at a router to support end‐to‐end congestion control in the Internet. Random early detection (RED) (IEEE/ACM Trans. Networking 1993; 1(4):397–413) enhanced the two functions by introducing queue length averaging and probabilistic early packet dropping. In particular, RED uses an exponentially weighted moving average (EWMA) queue length not only to detect incipient congestion but also to smooth the bursty incoming traffic and its resulting transient congestion. Following RED, many active queue management (AQM)‐based extensions have been proposed. However, many AQM proposals have shown severe problems with detection and control of the incipient congestion adaptively to the dynamically changing network situations. In this paper, we introduce and analyse a feedback control model of TCP/AQM dynamics. Then, we propose the Pro‐active Queue Management (PAQM) mechanism, which is able to provide proactive congestion avoidance and control using an adaptive congestion indicator and a control function under a wide range of traffic environments. The PAQM stabilizes the queue length around the desired level while giving smooth and low packet loss rates and high network resource utilization. Copyright © 2004 John Wiley & Sons, Ltd.

Design and analysis of multi-level active queue management mechanisms for emergency traffic

Multiple average-multiple threshold (MAMT) active queue management (AQM) is proposed as a solution for providing available and dependable service to traffic from emergency users after disasters. MAMT is a simple but effective approach that can be applied at strategic network locations, where heavy congestion is anticipated. It can provide low loss to emergency packets while dropping non-emergency packets only as much as necessary. Fluid flow analysis and simulation is conducted to provide guidelines for proper MAMT design, especially regarding the queue size and averaging parameters that are most important. This work considers non-responsive traffic exclusively, since non-responsive traffic types are currently getting the most attention from emergency management organizations. Plus, very little work has been performed regarding AQM and non-responsive traffic. It demonstrates queue oscillation problems that previously may have been attributed to the interactions between TCP and AQM, but which are actually inherent to AQM and can be greatly reduced with proper parameter settings. MAMT is shown to perform well over a range of loads and can effectively protect emergency traffic from surges in non-emergency traffic.

Nonlinear analysis of RED––a comparative study

Random Early Detection (RED) is an active queue management (AQM) mechanism for routers on the Internet. In this paper, performance of RED and Adaptive RED are compared from the viewpoint of nonlinear dynamics. In particular, we reveal the relationship between the performance of the network and its nonlinear dynamical behavior. We measure the maximal Lyapunov exponent and Hurst parameter of the average queue length of RED and Adaptive RED, as well as the throughput and packet loss rate of the aggregate traffic on the bottleneck link. Our simulation scenarios include FTP flows and Web flows, one-way and two-way traffic. In most situations, Adaptive RED has smaller maximal Lyapunov exponents, lower Hurst parameters, higher throughput and lower packet loss rate than that of RED. This confirms that Adaptive RED has better performance than RED.

Providing flow-based proportional differentiated services in class-based DiffServ routers

DiffServ provides class-based differentiated quality of service (QoS). For perception of users, flow-based service differentiation is much more important than class-based. Class-based QoS per-hop behaviour (PHB) is ensured in a DiffServ router. However, a flow in a higher class may get worse QoS than a flow in a lower class due to unbalanced traffic conditions. Therefore, a novel scheme with a Bloom-filter-based estimator for number of active flows, a dynamic weighted fair queueing (WFQ) scheduler, and a queue management mechanism to provide flow-based proportional QoS differentiation is proposed. This scheme dynamically adapts the WFQ weight of each class according to estimations of numbers of active flows in each class to provide proportional bandwidth differentiation for flows in different classes. Furthermore, a queue management mechanism with estimation of numbers of active flows and queue length feedback control is employed to control the queue length of each class around a target value to achieve delay differentiation. Through detailed simulations, the proposed scheme is shown to be able to provide flow-based proportional bandwidth and delay differentiations for flows in different classes at the same time.

Fair bandwidth sharing between unicast and multicast flows in best-effort networks

In this paper, we propose a simple scheduler called Service-Based Queuing (SBQ) to share the bandwidth fairly between unicast and multicast flows according to a new definition of fairness referred as the inter-service fairness. We also describe a new active queue management mechanism called Multicast Fair Queuing to fairly share the allowed multicast bandwidth among competing flows in the multicast queue. The simulation results obtained for very heterogeneous sources and links characteristics suggest that, on the one hand, SBQ achieves the expected aggregated bandwidth sharing among unicast and multicast service, and on the other hand the multicast flows remain TCP-friendly.

Load-tolerant differentiation with active queue management

Current work in the IETF aims at providing service differentiation on the Internet. One proposal is to provide loss differentiation by assigning levels of drop procedence to IP packets. In this paper, we evaluate the active queue management (AQM) mechanisms RED In and Out (RIO) and Weighted RED (WRED) in providing levels of drop precedence under different loads. For low drop precedence traffic, FIO and WRED can be configured to offer sheltering (i.e., low drop precedence traffic is protected from losses caused by higher drop precedence traffic). However, if traffic control fails or is inaccurate, such configurations can cause starvation of traffic at high drop precedence levels. Configuring WRED to instead offer relative differentiation can eliminate the risk of starvation. However, WRED cannot, without reconfiguration, both offer sheltering when low drop precedence traffic is properly controlled and avoid starvation at overload of low drop precedence traffic. To achieve this, we propose a new AQM mechanism, WRED with Thresholds (WRT). The benefit of WRT is that, without reconfiguration, it offers sheltering when low drop precedence traffic is properly controlled and relative differentiation otherwise. We present simulations showing that WRT has these properties.

Buffer management schemes for supporting TCP in gigabit routers with per-flow queueing

There has been much interest in using active queue management in routers in order to protect users from connections that are not very responsive to congestion notification. An Internet draft recommends schemes based on random early detection for achieving these goals, to the extent that it is possible, in a system without state. However, a stateless system with first-in/first-out (FIFO) queueing is very much handicapped in the degree to which flow isolation and fairness can be achieved. Starting with the observation that a stateless system is but one extreme in a spectrum of design choices and that per-flow queueing for a large number of flows is possible, we present active queue management mechanisms that are tailored to provide a high degree of isolation and fairness for TCP connections in a gigabit IP router using per-flow queueing. We show that IP flow state in a router can be bounded if the scheduling discipline used has finite memory, and we investigate the performance implications of different buffer management strategies in such a system. We show that merely using per-flow scheduling is not sufficient to achieve effective isolation and fairness, and it must be combined with appropriate buffer management strategies.