Optimizing DRE System Performance with the SMACK Cache Efficiency Metric

Distributed real-time and embedded (DRE) systems are often subject to stringent timing constraints. Scheduling techniques, such as rate monotonic scheduling, can be used to ensure that real-time deadlines are met. Although a processor cache can reduce the time required for a task schedule to execute, multiple task execution schedules may exist that meet deadlines but differ in cache utilization efficiency. It is hard to determine which task execution schedules will utilize the processor cache most efficiently and provide the greatest reductions in execution time without jeopardizing real-time deadlines. The work in this paper provides three key contributions to predictive performance evaluation of processor caching in DRE systems. First, we present the System Metric for Application Cache Knowledge (SMACK), which is a novel approach to quantify the expected cache utilization efficiency of different schedules. Second, we employ SMACK to predict the relative execution time and cache misses of 11 simulated software systems with 2 different execution schedules per system. Third, we empirically evaluate the impact of using SMACK as a heuristic to alter task schedules to reduce system execution time. Our results show that heuristic scheduling with SMACK increases cache performance, reduces execution time, and satisfies real-time scheduling constraints and safety requirements without requiring significant hardware or software changes. ∗This work was sponsored in part by the Air Force Research Lab.