Dynamic Arbitration of Memory Requests with TDM-like Guarantees
Résumé
A major challenge with multi-cores in real-time systems is contention between concurrent accesses to shared memory. Dynamic arbitration schemes allow for an optimal utilization of the system's memory, while sacrificing time predictability. Time-Division multiplexing (TDM), on the other hand, sacrifices average-case performance in favor of predictability. In this work, we explore a dynamic arbitration scheme that is essentially based on TDM, and thus preserves many of its guarantees.