Dynamic Incentives for Congestion Control

Halim

IEEE Commun. Surv. Tutorials

2016

Providing congestion control in the Internet, while ensuring fairness among myriad of heterogeneous flows is a challenging task. The conventional wisdom is to rely on end-user applications cooperatively deploying congestion control mechanisms to achieve high network utilization and some degree of fairness among flows. However, as the Internet has evolved to encompass all of society, such a cooperative behavior from enduser applications is not always granted. Applications may simply act selfishly to be more competitive through bandwidth abuse. Bandwidth starvation may also arise unintentionally depending on the nature of traffic sources. The ensuing impact can be severe fairness hazard and even congestion collapse. Router-based queue management schemes driven by fairness objectives thus become an inescapable necessity for fairly sharing network resources.Given a significant volume of literature relating to fairnessdriven queue management schemes, there has remained a need for a broader and coherent survey. This paper presents a systematic and comprehensive review of eminent fairness-driven queue management schemes from the inception of the concept and the preliminary work to the most recent work. We present a new taxonomy of categorizing fairness-driven queue management schemes. We discuss design approaches and key attributes of these schemes and provide their comparison and analysis. Based on the outcomes of this survey, we discuss a number of open issues and provide generic design guidelines and future directions for the research in this field. Fairness-Driven Queue Management: A Survey and Taxonomy 2 c packets/s Scheduling algorithm Dequeueend B packets Buffer k l 1 l Enqueueend Output link Input links Queue management algorithm mechanisms at all, either deliberately or by accident, and generate misbehaving traffic [9]. Furthermore, there are applications that selfishly improve throughput by splitting a single TCP connection into multiple connections [10], [3]. All such applications and the resulting traffic aggregates are potentially dangerous as they inflict unfairness in the network, and can eventually cause congestion collapse [9], [11].In response to these problems, there has been a long history, dating back to [12], of realizing that routers play a significant role in fair bandwidth allocation. The technical report by Floyd and Fall [13] (later published as [3]) is, however, the first one to extensively demonstrate the danger of unfairness due to unresponsive flows. Floyd and Fall [13] argue that the incentives for cooperative behavior can only come from the network itself, and therefore, routers inevitably need to deploy mechanisms to provide an incentive structure for applications to use end-to-end congestion control. The report also proposes queue management based techniques for identifying and restricting unresponsive flows. Since then, a number of fairness-driven queue management schemes have been proposed to shield responsive flows and to regulate unresponsive and aggressive flows.This paper provid...

Section: Providing the Right Incentivesmentioning

confidence: 99%

Fairness-Driven Queue Management: A Survey and Taxonomy

Halim

IEEE Commun. Surv. Tutorials

2016

2019 IEEE 58th Conference on Decision and Control (CDC)

“…Research in aggregative games has surged in recent years due to their suitability to model decision problems in various application domains: from demand-side management for the smart grid, [1] and for electric vehicles, [2], [3], to wireless systems, [4], network congestion control, [5] and competitive markets, [6]. Aggregative games are non-cooperative games in which each (player's) agent's cost depends on some aggregate effect of all other agents' actions.…”

Section: Introductionmentioning

confidence: 99%

Distributed GNE seeking over networks in aggregative games with coupled constraints via forward-backward operator splitting

Gadjov

Pavel

2019

We consider the framework of aggregative games with affine coupling constraints, where agents have partial information of the aggregate value and can only obtain information from neighbouring agents. We propose a single-layer distributed algorithm that reaches a variational generalized Nash equilibrium, under constant step sizes. The algorithm works on a single timescale, i.e., does not require multiple communication rounds between agents before updating their action. The convergence proof leverages an invariance property of the aggregate estimates and relies on a forward-backward splitting for two preconditioned operators and their restricted (strong) monotonicity properties on the consensus subspace.

Intl J Robust & Nonlinear

“…Aggregative games play an important role in many applications, such as communication networks, smart grids, and market economics . In aggregative games, the cost function of every player relies on not only its own decision variable but also the aggregate of the decision of all players.…”

Section: Introductionmentioning

confidence: 99%

Distributed algorithm design for aggregative games of disturbed multiagent systems over weight‐balanced digraphs

Deng

Nian

2018

Summary In this paper, an aggregative game of multiagent systems over weight‐balanced digraphs is studied, where the decisions of all players are coupled by linear constraints. Different from the well‐known aggregative games, the dynamics of players are disturbed first‐order linear systems in our problem. In order to seek the variational generalized Nash equilibrium (GNE) of the game, a distributed algorithm is developed via gradient descent, internal model, and dynamic average consensus, where the gradient is for seeking the variational GNE, the internal model is for rejecting the exogenous disturbances, and the dynamic average consensus is for estimating the aggregate of the decisions of all players. The exponential convergence of the algorithm to the variational GNE is analyzed by constructing suitable Lyapunov functions. Finally, the effectiveness of our method is illustrated via two examples.