Bandwidth auctions and their parallels to power
Abstract:
This talk will present recent contributions to auctions for network bandwidth, and discuss connections to auctions of power.
In a bandwidth auction, a network operator sells its transmission capacity to bidders interested in a certain service bandwidth between two endpoints. The operator must then allocate its resources to maximize revenue. Over a single time interval, this is an integer program susceptible of tractable approximations; in particular, variants of the network utility maximization framework can be used to obtain distributed solutions. However, an additional issue that must be accommodated are inter-temporal constraints: successful bidders should have a reserved service over multiple auction periods, which makes the optimization of revenue a dynamic optimization problem. We show how a receding horizon method provides tractable approximations to its solution.
We will comment on similarities and differences with power auctions. The latter are often double-sided, with bids and asks for demand and supply; they are simpler in that they allocate a single commodity, but have more nonlinear constraints stemming from the AC power flow. Inter-temporal constraints do appear naturally on the supply side, where generators have minimum service times. We will give preliminary thoughts on possible research directions in this active area.