Correction of an Augmentation Bound Analysis for Parallel Real-Time Tasks

With the advent of multi-core processors, real-time scheduling of parallel tasks has received considerable attention in recent years. In this paper, we propose some significant corrections in a recent work of Lakshmanan et al. on parallel task scheduling. Lakshmanan et al. has proposed a transformation of parallel tasks into sequential tasks, and has claimed a resource augmentation bound of 3.42 for partitioned deadline monotonic (DM) scheduling of the transformed tasks. In this paper, we demonstrate that their resource augmentation bound is incorrect. We also demonstrate that no practical resource augmentation bound can be derived based on their proposed task transformation. We therefore propose a different technique for task transformation that requires a resource augmentation bound of 5 for partitioned DM scheduling. Type of Report: Other Department of Computer Science & Engineering Washington University in St. Louis Campus Box 1045 St. Louis, MO 63130 ph: (314) 935-6160 Noname manuscript No. (will be inserted by the editor) Correction of an Augmentation Bound Analysis for Parallel Real-Time Tasks Abusayeed Saifullah · Kunal Agrawal · Chenyang Lu · Christopher Gill Received: date / Accepted: date Abstract With the advent of multi-core processors, real-time scheduling of parallel tasks has received considerable attention in recent years. In this paper, we propose some significant corrections in a recent work of Lakshmanan et al. on parallel task scheduling. Lakshmanan et al. has proposed a transformation of parallel tasks into sequential tasks, and has claimed a resource augmentation bound of 3.42 for partitioned deadline monotonic (DM) scheduling of the transformed tasks. In this paper, we demonstrate that their resource augmentation bound is incorrect. We also demonstrate that no practical resource augmentation bound can be derived based on their proposed task transformation. We therefore propose a different technique for task transformation that requires a resource augmentation bound of 5 for partitioned DM scheduling.With the advent of multi-core processors, real-time scheduling of parallel tasks has received considerable attention in recent years. In this paper, we propose some significant corrections in a recent work of Lakshmanan et al. on parallel task scheduling. Lakshmanan et al. has proposed a transformation of parallel tasks into sequential tasks, and has claimed a resource augmentation bound of 3.42 for partitioned deadline monotonic (DM) scheduling of the transformed tasks. In this paper, we demonstrate that their resource augmentation bound is incorrect. We also demonstrate that no practical resource augmentation bound can be derived based on their proposed task transformation. We therefore propose a different technique for task transformation that requires a resource augmentation bound of 5 for partitioned DM scheduling.