Reverse auction-based job assignment among foundry fabs

Motivated by foundry service provisioning in the semiconductor industry, this paper adopts a reverse auction-based mechanism to model job order assignment from a job owner (auctioneers) to a few qualified and competing foundry fabs (bidders). A job order owner announces job requirements and payments for fabs to bid on while qualified fabs bid on a job by offering the discount to payment and processing schedule of the job. This model aims at capturing the noncooperative gaming among the job owner and the fabs because of the private information such as objectives, valuation of jobs, available capacity and constraints. Two integer-programming formulations are formulated for the reverse auction of job assignment; one is deterministic and models a common practice of mean value-based bidder decision-making while the other is stochastic and captures bidders' consideration of uncertainties and the associated risk in decision-making. A Lagrange relaxation-based, near-optimal scheduling method is developed to model a bidder's selection of job-to-bid and schedule for the deterministic formulation. The bidder's job selection and scheduling model for the stochastic formulation combines simulation and simple heuristics. A bidder decides the discount offer for each job-to-bid by a simple fixed-increment scheme. The auctioneer simply assigns the job to the bid with the highest discount offer in each round of bidding. Analyses show that the reverse auction model leads to an equilibrium solution, where no single bidder would unilaterally deviate from the auction result. Numerical study by using the reverse auction model demonstrates that consideration of uncertainties in bidders' decisions has a larger impact on performance of both bidders and the auctioneer than optimality of the mean value-based bidder scheduling algorithm.