A private good/public good decomposition for optimal flow control of anM/M/1queue

Flow control for an
M/M/1
queue with several users conceptually consists of two parts: determining the total throughput (and thus mean delay) and determining the rate allocated to each individual user. An incentive compatible hill climbing algorithm that determines the optimal rate allocation is presented. It is shown that the algorithm can be viewed as a public good algorithm that determines the total, and a private good algorithm that determines the allocation, operating simultaneously.