2001ACM SIGCOMM Computer Communication ReviewRequires access

Aggregate traffic performance with active queue management and drop from tail

Gianluca Iannaccone, Martin May, Christophe Diot

Open publisher page 100 citations

Abstract

Active queue management (AQM) refers to a family of packet dropping mechanisms for router queues that has been proposed to support end-to-end congestion control mechanisms in the Internet. In this paper, we examine the performance of AQM mechanisms by varying two parameters: the queue size and the dropping function. AQM flavors considered include "RED", the more recently proposed "Gentle RED" and an additional mechanism we call "Gentle RED with instantaneous queue size".We use experimentation to analyze the performance of the AQM mechanisms identified above on the aggregate traffic going through a congested router. The metrics used are: TCP goodput, TCP and UDP loss rate, queueing delay and consecutive loss probability. The AQM mechanisms are compared to Drop from Tail, the buffer management mechanism currently found in most operational routers.The major observation is that AQM mechanisms have a minor impact on the aggregate performance metrics we observe. On the other hand, we observe an important sensitivity of the AQMs considered to traffic characteristics that may compromise their operational deployment.

About this research paper

What this paper is about

Active queue management (AQM) refers to a family of packet dropping mechanisms for router queues that has been proposed to support end-to-end congestion control mechanisms in the Internet. In this paper, we examine the performance of AQM mechanisms by varying two parameters: the queue size and the dropping function. AQM flavors considered include "RED", the more recently proposed "Gentle RED" and an additional mechanism we call "Gentle RED with instantaneous queue size".We use experimentation to analyze the performance of the AQM mechanisms identified above on the aggregate traffic going through a congested router. The metrics used are: TCP goodput, TCP and UDP loss rate, queueing delay and consecutive loss probability. The AQM mechanisms are compared to Drop from Tail, the buffer management mechanism currently found in most operational routers.The major observation is that AQM mechanisms have a minor impact on the aggregate performance metrics we observe. On the other hand, we observe an important sensitivity of the AQMs considered to traffic characteristics that may compromise their operational deployment.

Why it matters

OpenAlex reports 100 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

Active queue management (AQM) refers to a family of packet dropping mechanisms for router queues that has been proposed to support end-to-end congestion control mechanisms in the Internet. In this paper, we examine the performance of AQM mechanisms by varying two parameters: the queue size and the dropping function. AQM flavors considered include "RED", the more recently proposed "Gentle RED" and an additional mechanism we call "Gentle RED with instantaneous queue size".We use experimentation to analyze the performance of the AQM mechanisms identified above on the aggregate traffic going through a congested router. The metrics used are: TCP goodput, TCP and UDP loss rate, queueing delay and consecutive loss probability. The AQM mechanisms are compared to Drop from Tail, the buffer management mechanism currently found in most operational routers.The major observation is that AQM mechanisms have a minor impact on the aggregate performance metrics we observe. On the other hand, we observe an important sensitivity of the AQMs considered to traffic characteristics that may compromise their operational deployment.

Key concepts: Active queue management, Goodput, Random early detection, Computer science, Computer network, Router, Queueing theory, Queue

Related papers

Back to paper searchBrowse research topicsOriginal source
Aggregate traffic performance with active queue management and drop from tail — Research Paper | ScholarLens