Distributed algorithms for english auctions

Abstract

This paper deals with the Implementation of an English auction on a distributed system. We assume that all messages are restricted to bids and resignations (referred to as the limited communication assumption) and that all participants are trying to maximize there gains (referred to as the prudence assumption). We also assume that bidders are risk-neutral, and that the underlying communication network Is complete, asynchronous and failure-free. Under these assumptions, we show that the time and communication requirements of any auction process are Ω(M2) and Ω(M2 + n) respectively, where M2 denotes the second largest valuation of a participant In the auction. We then develop a number of distributed algorithmic Implementations for English auction, analyze their time and communication requirements, and propose an algorithm achieving optimal time and communication i.e., meeting the above lower bounds.Finally we discuss extensions to the case of dynamically joining participants.