Predictable Memory-CPU Co-Scheduling with Support for Latency-Sensitive Tasks