LookAhead: The Optimal Non-decreasing Index Policy for a Time-Varying Holding Cost problem
By: Keerthana Gurushankar , Zhouzi Li , Mor Harchol-Balter and more
In practice, the cost of delaying a job can grow as the job waits. Such behavior is modeled by the Time-Varying Holding Cost (TVHC) problem, where each job's instantaneous holding cost increases with its current age (a job's age is the time since it arrived). The goal of the TVHC problem is to find a scheduling policy that minimizes the time-average total holding cost across all jobs. However, no optimality results are known for the TVHC problem outside of the asymptotic regime. In this paper, we study a simple yet still challenging special case: A two-class M/M/1 queue in which class 1 jobs incur a non-decreasing, time-varying holding cost and class 2 jobs incur a constant holding cost. Our main contribution is deriving the first optimal (non-decreasing) index policy for this special case of the TVHC problem. Our optimal policy, called LookAhead, stems from the following idea: Rather than considering each job's current holding cost when making scheduling decisions, we should look at their cost some $X$ time into the future, where this $X$ is intuitively called the ``lookahead amount." This paper derives that optimal lookahead amount.
Similar Papers
Scheduling with Uncertain Holding Costs and its Application to Content Moderation
Data Structures and Algorithms
Helps social media remove bad posts faster.
Improving Upon the generalized c-mu rule: a Whittle approach
Performance
Schedules jobs to save money by reducing waiting time.
Queues with inspection cost: To see or not to see?
CS and Game Theory
Helps people decide when to wait or pay to see.