Cache Partitioning

Context: pre-emptively scheduled multi-task systems with caches Idea: each task is assigned a dedicated part of the cache Pros: I predictable system I less dynamic behaviour I standard WCET analysis applicable Cons: I higher (non-preempted) execution times I implementation overhead Commonly accepted assumption: (i) Cache partitioning increases predictability (ii) Predictability is good for real-time systems ⇒ Cache partitioning is good! Evaluation of Cache Partitioning for Hard Real-Time Systems 2 / 30 We want to know: I How useful is cache partitioning? I When to use cache partitioning? I Which system/task parameters influence usefulness? Differences to related work: 1. based on analytical approaches (as used for safety-critical systems) I instead of measurements-based evaluation [3] or simplified execution time models [4] 2. using optimal cache partitioning (wrt. schedulability) I instead of focussing on implementation details [8, 10, 11] 3. comparison with state-of-the-art pre-emption cost analysis I instead of comparisons with systems without caches [12] ∗ Terms and conditions may appl