Capacity Management in Decentralized Networks

Bottleneck analysis is a useful tool in capacity planning for centrally controlled network systems. However, under a decentralized network where individual users are allowed to select their own routes, straightforward application of bottleneck analysis does not necessarily yield an optimal performance. It may even hurt the system performance—an aspect of Braess's paradox. We investigate the capacity expansion problem for a decentralized system with general network topology. To this end, we first discuss the short-run problem and show that the externality pricing solves the joint problem of demand and routing control. We then study the capacity expansion/reduction problem for decentralized systems that may or may not be optimally controlled in the short run.