Transient traffic congestion control with Traveling Auctions

The use of market mechanisms to solve computer science problems such as resource sharing, load distribution and network routing, is gaining significant traction. In this paper, we investigate new market mechanisms to solve the problem of bandwidth sharing in wireless networks for transient traffic congestion often generated by event-driven packet flows. Typically, such congestion is transient at each node it arises since the bursts of data move following the event. We first demonstrate that a previously proposed strategy that greedily selects winners in repeated routing auctions is not globally optimal in such a case. We also demonstrate, by evaluating a lookahead mechanism for winner selection in the corresponding auctions, that greedy algorithm approximates optimal selection very closely. Then, we introduce and evaluate a novel mechanism which we call Traveling Auctions to address the problem of transient congestion. We experimentally show that using Traveling Auctions mechanism improves the network performance.