Job-Level Dynamic Cache Allocation for Real-Time Systems under Shared LLC Interference

ACM Transactions on Embedded Computing Systems. Efficient management of shared cache resources is a key challenge in multicore real-time systems. Mechanisms such as Intel’s Cache Allocation Technology (CAT) enable software control over last-level cache (LLC) partitions, providing opportunities to reduce interference and tighten worst-case execution time (WCET) bounds. Existing approaches typically rely on static cache allocation, which can lead to over-provisioning and inefficient cache usage. This article introduces a job-level dynamic cache allocation strategy for multicore systems in which critical real-time tasks execute on a dedicated core while sharing the LLC with best-effort workloads executing on the remaining cores. We formulate the allocation problem as a constraint-based optimization model that simultaneously enforces schedulability constraints, cache capacity limits, and preemption-induced concurrency. By allocating and releasing cache partitions at job boundaries, our approach adapts to temporal variations in job activity, minimizing the cache dedicated to critical tasks while preserving their schedulability. Through evaluation on synthetic task sets, we show that job-level dynamic allocation improves schedulability and systematically increases the cache available to non-critical tasks compared to task-level static or dynamic strategies. Moreover, it often requires fewer allocation operations, indicating that increased flexibility does not entail higher runtime overhead. These results demonstrate that job-level cache allocation is a promising direction for efficient and predictable resource management in multicore real-time systems.