Routing and wavelength assignment in optical networks

The problem of routing and wavelength assignment (RWA) is critically important for increasing the efficiency of wavelength-routed all-optical networks. Given the physical network structure and the required connections, the RWA problem is to select a suitable path and wavelength among the many possible choices for each connection so that no two paths sharing a link are assigned the same wavelength. In work to date, this problem has been formulated as a difficult integer programming problem that does not lend itself to efficient solution or insightful analysis. In this work, we propose several novel optimization problem formulations that offer the promise of radical improvements over the existing methods. We adopt a (quasi-)static view of the problem and propose new integer-linear programming formulations, which can be addressed with highly efficient linear (not integer) programming methods and yield optimal or near-optimal RWA policies. The fact that this is possible is surprising, and is the starting point for new and greatly improved methods for RWA. Aside from its intrinsic value, the quasi-static solution method can form the basis for suboptimal solution methods for the stochastic/dynamic settings.     

Distributed Subgradient Methods for Multi-Agent Optimization

We study a distributed computation model for optimizing a sum of convex objective functions corresponding to multiple agents. For solving this (not necessarily smooth) optimization problem, we consider a subgradient method that is distributed among the agents. The method involves every agent minimizing his/her own objective function while exchanging information locally with other agents in the network over a time-varying topology. We provide convergence results and convergence rate estimates for the subgradient method. Our convergence rate results explicitly characterize the tradeoff between a desired accuracy of the generated approximate optimal solutions and the number of iterations needed to achieve the accuracy.     

Competition in Parallel-Serial Networks

We study the efficiency implications of competition among profit-maximizing service providers in communication networks. Service providers set prices for transmission of flows through their (sub)network. The central question is whether the presence of prices will help or hinder network performance. We investigate this question by considering the difference between users' willingness to pay and delay costs as the efficiency metric. Previous work has demonstrated that in networks consisting of parallel links, efficiency losses from competition are bounded. Nevertheless, parallel-link networks are special, and in most networks, traffic has to simultaneously traverse links (or subnetworks) operated by independent service providers. The simplest network topology allowing this feature is the parallel-serial structure, which we study in this paper. In contrast to existing results, we show that in the presence of serial links, the efficiency loss relative to the social optimum can be arbitrarily large. The reason for this degradation of performance is the double marginalization problem, whereby each serial provider charges high prices not taking into account the effect of this strategy on the profits of other providers along the same path. Nevertheless, when there are no delay costs without transmission (i.e., latencies at zero are equal to zero), irrespective of the number of serial and parallel providers, the efficiency of strong oligopoly equilibria can be bounded by 1/2, where strong oligopoly equilibria are equilibria in which each provider plays a strict best response and all of the traffic is transmitted. However, even with strong oligopoly equilibria, inefficiency can be arbitrarily large when the assumption of no delay costs without transmission is relaxed.     

Efficiency and Braess' Paradox under pricing in general networks

We study the flow control and routing decisions of self-interested users in a general congested network where a single profit-maximizing service provider sets prices for different paths in the network. We define an equilibrium of the user choices. We then define the monopoly equilibrium (ME) as the equilibrium prices set by the service provider and the corresponding user equilibrium. We analyze the networks containing different types of user utilities: elastic or inelastic. For a network containing inelastic user utilities, we show the flow allocations at the ME and the social optimum are the same. For a network containing elastic user utilities, we explicitly characterize the ME and study its performance relative to the user equilibrium at 0 prices and the social optimum that would result from centrally maximizing the aggregate system utility. We also define Braess' Paradox for a network involving pricing and show that Braess' Paradox does not occur under monopoly prices.     

Partially Optimal Routing

Most large-scale communication networks, such as the Internet, consist of interconnected administrative domains. While source (or selfish) routing, where transmission follows the least cost path for each source, is reasonable across domains, service providers typically engage in traffic engineering to improve operating performance within their own network. Motivated by this observation, we develop and analyze a model of partially optimal routing, where optimal routing within subnetworks is overlaid with selfish routing across domains. We demonstrate that optimal routing within a subnetwork does not necessarily improve the performance of the overall network. In particular, when Braess' paradox occurs in the network, partially optimal routing may lead to worse overall network performance. We provide bounds on the worst-case loss of efficiency that can occur due to partially optimal routing. For example, when all congestion costs can be represented by affine latency functions and all administrative domains have a single entry and exit point, the worst-case loss of efficiency is no worse than 25% relative to the optimal solution. In the presence of administrative domains incorporating multiple entry and/or exit points, however, the performance of partially optimal routing can be arbitrarily inefficient even with linear latencies. We also provide conditions for traffic engineering to be individually optimal for service providers.