Abstract Performance Isolation for Mixed Criticality Real-time System on Multicore with Xen Hypervisor

Multicore processors have imported the powerful computing capacity to real-time systems, allowing the multi-task execution on the co-running cores. Meanwhile, the competition for the shared resources among the on-die cores brings the side-effects which highly degrades the performance of single cores and puts the real-time requirements in danger. This thesis work is focused on addressing the memory access contentions on the real-time systems with mixed-criticality. A throttling algorithm is designed to control the memory access flows from the interfering side, securing the deadlines of critical tasks. We implemented the throttling framework on Xen Hypervisor and evaluated the overall isolating performance with a set of benchmarks. The results prove the effectiveness of our design