Abstract Selfish routing is a classical mathematical model of how self-interested users might route traffic through a congested network. The outcome of selfish routing is generally inefficient, in that it fails to optimize natural objective functions. The price of anarchy is a quantitative measure of this inefficiency. We survey recent work that analyzes the price of anarchy of selfish routing. We also describe related results on bounding the worst-possible severity of a phenomenon called Braess's Paradox, and on three techniques for reducing the price of anarchy of selfish routing. This survey concentrates on the contributions of the author's PhD thesis, but also discusses several more recent results in the area.

We study the negative consequences of selfish behavior in a congested network and economic means of influencing such behavior. We consider the model of selfish routing defined by Wardrop [30] and studied in a computer science context by Roughgarden and Tardos [26]. In this model, the latency experienced by network traffic on an edge of the network is a function of the edge congestion, and network users are assumed to selfishly route traffic on minimum-latency paths. The quality of a routing of traffic is measured by the sum of travel times (the total latency).

We introduce the notion of coordination mechanisms to improve the performance in systems with independent selfish and non-colluding agents. The quality of a coordination mechanism is measured by its price of anarchy—the worst-case performance of a Nash equilibrium over the (centrally controlled) social optimum. We give upper and lower bounds for the price of anarchy for selfish task allocation and congestion games.

Mechanism design is now a standard tool in computer science for aligning the incentives of self-interested agents with the objectives of a system designer. There is, however, a fundamental disconnect between the traditional application domains of mechanism design (such as auctions) and those arising in computer science (such as networks): while monetary transfers (i.e., payments) are essential for most of the known positive results in mechanism design, they are undesirable or even technologically infeasible in many computer systems. Classical impossibility results imply that the reach of mechanisms without transfers is severely limited. Computer systems typically do have the ability to reduce service quality—routing systems can drop or delay traffic, scheduling protocols can delay the release of jobs, and computational payment schemes can require computational payments from users (e.g., in spam-fighting systems). Service degradation is tantamount to requiring that users burn money, and such “payments ” can be used to influence the preferences of the agents at a cost of degrading the social surplus. We develop a framework for the design and analysis of money-burning mechanisms to maximize the residual surplus— the total value of the chosen outcome minus the payments required. Our primary contributions are the following. • We define a general template for prior-free optimal mechanism design that explicitly connects Bayesian optimal mechanism design, the dominant paradigm in economics, with worst-case analysis. In particular, we establish a general and principled way to identify appropriate performance benchmarks in prior-free mechanism design. • For general single-parameter agent settings, we char-

Abstract not found

We investigate adaptive routing policies for large networks in which agents reroute traffic based on old information. It is a well known and practically relevant problem that old information can lead to undesirable oscillation effects resulting in poor performance. We investigate how adaptive routing policies should be designed such that these effects can be avoided. The network is represented by a general graph with latency functions on the edges. Traffic is managed by a large number of agents each of which is responsible for a negligible amount of traffic. Initially the agents ’ routing paths are chosen in an arbitrary fashion. From time to time each agent revises her routing strategy by sampling another path and switching with positive probability to this path if it promises smaller latencies. As the information on which the agent bases her decision might be stale, however, this does not necessarily lead to an improvement. The points of time at which agents revise their strategy are generated by a Poisson distribution. Stale information is modelled in form of a bulletin board that is updated periodically and lists the latencies on all edges. We analyze such a distributed routing process in the socalled fluid limit, that is, we use differential equations describing the fractions of traffic on different paths over time. In our model, we can show the following effects. Simple routing policies that always switch to the better alternative lead to oscillation, regardless at which frequency the bulletin board is updated. Oscillation effects can be avoided, however, when using smooth adaption policies that do not always switch to better alternatives but only with a probability depending on the advantage in the latency. In fact, such policies have dynamics that converge to a fixed point corresponding to a Nash equilibrium for the underlying routing game, provided the update periods are not too large.

We investigate the influence of different algorithmic choices on the approximation ratio in selfish scheduling. Our goal is to design local policies that minimize the inefficiency of resulting equilibria. In particular, we design optimal coordination mechanisms for unrelated machine scheduling, and improve the known approximation ratio from Θ(m) to Θ(log m), where m is the number of machines. A local policy for each machine orders the set of jobs assigned to it only based on parameters of those jobs. A strongly local policy only uses the processing time of jobs on the the same machine. We prove that the approximation ratio of any set of strongly local ordering policies in equilibria is at least Ω(m). In particular, it implies that the approximation ratio of a greedy shortest-first algorithm for machine scheduling is at least Ω(m). This closes the gap between the known lower and upper bounds for this problem, and answers an open question raised by Ibarra and Kim [16], and Davis and Jaffe [10]. We then design a local ordering policy with the approximation ratio of Θ(log m) in equilibria, and prove that this policy is optimal among all local ordering policies. This policy orders the jobs in the non-decreasing order of their inefficiency, i.e, the ratio between the processing time on that machine over the minimum processing time. Finally, we show that best responses of players for the inefficiency-based policy may not converge to a pure Nash equilibrium, and present a Θ(log² m) policy for which we can prove fast convergence of best responses to pure Nash equilibria.

Abstract not found

In large-scale communication networks, like the Internet, it is usually impossible to globally manage network traffic. In the absence of global control it is typically assumed in traffic modeling that network users follow the most rational approach, that is, they behave selfishly to optimize their own individual welfare. This motivates the analysis of network traffic using models from Game Theory in which each player is aware of the situation facing all other players and tries to minimize its cost. Under these assumptions, the routing process should arrive into a so-called Nash equilibrium in which no network user has an incentive to change its strategy. It is well known (and easy to see) that Nash equilibria do not always optimize the overall performance of the system. In this survey we investigate the relation between these two notions for traffic routing: network performance in the Nash equilibria and the optimal performance of the system. Our main focus is on the analysis of the coordination ratio, which is the ratio between the worst possible Nash equilibrium and the overall optimum. In other words, this analysis seeks the price of uncoordinated

In this paper, we study the price of anarchy of traffic routing, under the assumption that users are partially altruistic or spiteful. We model such behavior by positing that the “cost ” perceived by a user is a linear combination of the actual latency of the route chosen (selfish component), and the increase in latency the user causes for others (altruistic component). We show that if all users have a coefficient of at least β> 0 for the altruistic component, then the price of anarchy is bounded by 1/β, for all network topologies, arbitrary commodities, and arbitrary semi-convex latency functions. We extend this result to give more precise bounds on the price of anarchy for specific classes of latency functions, even for β < 0 modeling spiteful behavior. In particular, we show that if all latency functions are linear, the price of anarchy is bounded by 4/(3 + 2β − β 2). We next study non-uniform altruism distributions, where different users may have different coefficients β. We prove that all such games, even with infinitely many types of players, have a Nash Equilibrium. We show that if the average of the coefficients for the altruistic components of all users is ¯ β, then the price of anarchy is bounded by 1 / ¯ β, for single commodity parallel link networks, and arbitrary convex latency functions. In particular, this result generalizes, albeit non-constructively, the Stackelberg routing results of Roughgarden and of Swamy. More generally, we bound the price of anarchy based on the class of allowable latency functions, and as a corollary obtain tighter bounds for Stackelberg routing than a recent result of Swamy.