Approximating the Performance of a Batch Service Queue Using the M/Mk/1 Model

Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The published version is available at: Batching plays an important role in semiconductor fabs, and it can lead to inefficiency if not treated with care. The performance of parallel batch processes is often approximated by the G/G/1 based approximate models. By carefully examining the existing models, the dependence between queueing time and wait-to-batch time has been identified. A new improved model for parallel batch systems is proposed to exploit this dependence. The computation of the new model is still simple and fast, but it gives better approximation by reducing the systematic error in earlier models which ignored the dependence between queueing time and wait-to-batch time. Note to Practitioners: Practical manufacturing systems are usually very complex. In order to analyze system behavior, assumptions are made to enable the analysis. A decomposition approach is commonly adopted based on an independence assumption. However, any two components in a complex manufacturing system are seldom completely independent. When there are real dependencies, we should take special care with analyses based on decomposition. This paper examines parallel batch processing and through detailed analysis, identifies a particular dependence and proposes an improved model that is validated through experimentation.