Description | Paccar Hall, Room 293 Seminar Speaker: Leela Nageswaran Affiliation: Carnegie Mellon University Area: Operations Management Name of Presentation: Queues with Redundancy: Is Waiting in Multiple Lines Fair? Abstract: In a service system, a redundant customer is one who may join multiple queues simultaneously and is "served" when any one of her copies completes service; systems with redundant customers range from multiple listing for organ transplants to supermarkets with multiple checkout lines. By considering two queues serving two classes of customers, one of which is redundant, our model provides fundamental insights on fairness, optimal queue-joining policies, and the value of information in such systems, from the non-redundant customers' perspective. Specifically, we prove that non-redundant customers forming independent Poisson streams to each queue actually benefit under redundancy of the other class compared to when the other class joins the shortest queue (JSQ) if the queues are symmetric (i.e., in this case redundancy is fair); however, there are situations where they are worse off if the queues are asymmetric. In this latter case redundancy may be seen as unfair. We also prove that JSQ does not always result in the lowest delay for a non-redundant customer if complete system state information (including which customers are redundant) is available, but that JSQ is optimal if only queue length information is observable. We extend several of our results to include unequal service rates and customer abandonments, and to systems with more than two servers. |
---|