Performance isolation for real-time systems with Xen hypervisor on multi-cores

Virtualization techniques are gaining significant interests in embedded real-time system design. However, existing virtualization platforms lack strong performance isolation among virtual machines. In this work we propose a method to monitor and control the shared memory accesses of individual virtual machines on multi-core processors with Xen hypervisor, to enhance the performance isolation among virtual machines and improve the timing predictability of real-time applications. Experiments with the SPEC2006 benchmark programs are conducted to validate the proposed method.

[1]  Lui Sha,et al.  Real-Time Control of I/O COTS Peripherals for Embedded Systems , 2009, 2009 30th IEEE Real-Time Systems Symposium.

[2]  Chenyang Lu,et al.  RT-Xen: Towards real-time hypervisor scheduling in Xen , 2011, 2011 Proceedings of the Ninth ACM International Conference on Embedded Software (EMSOFT).

[3]  Gernot Heiser,et al.  The role of virtualization in embedded systems , 2008, IIES '08.

[4]  Alexandra Fedorova,et al.  Addressing shared resource contention in multicore processors via scheduling , 2010, ASPLOS XV.

[5]  Kees G. W. Goossens,et al.  Architectures and modeling of predictable memory controllers for improved system integration , 2011, 2011 Design, Automation & Test in Europe.

[6]  Luca Benini,et al.  Combining Simulation and Formal Methods for System-Level Performance Analysis , 2006, Proceedings of the Design Automation & Test in Europe Conference.

[7]  Michael D. Smith,et al.  Improving Performance Isolation on Chip Multiprocessors via an Operating System Scheduler , 2007, 16th International Conference on Parallel Architecture and Compilation Techniques (PACT 2007).

[8]  Lothar Thiele,et al.  A general framework for analysing system properties in platform-based embedded system designs , 2003, 2003 Design, Automation and Test in Europe Conference and Exhibition.

[9]  Godmar Back,et al.  Perfctr-Xen: a framework for performance counter virtualization , 2011, VEE '11.