Cache-Aware Real-Time Scheduling on Multicore Platforms

Cache-Aware Real-Time Scheduling on Multicore Platforms

Abstract:

Multicore architectures, which have multiple processing units on asingle chip, have been adopted by most chip manufacturers. Most suchchips contain on-chip caches that are sharedby some or all of the cores on the chip. To efficiently use theavailable processing resources on such platforms, scheduling methodsmust be aware of these caches. In this talk, I will present a method forimproving cache performance when scheduling real-time workloads.Additionally, I will discuss our ongoing work on methods to dynamicallyprofile the cache behavior of real-time tasks, which allows ourscheduling method to be effectively employed. These scheduling andprofiling methods are especially applicable when multiple multithreadedreal-time applications exist with large working sets. As this couldeasily be the case for a multimedia server, we also present apreliminary case study that shows how our best-performing heuristics canimprove the end-user performance of video encoding applications --- weplan to expand this study in future work.