Abstract: Xen does not consider the impact of CPU intermittent faults. On the basis of this, this paper builds a time model for simulating intermittent fault in Xen. In the model, domains in unmodified Xen system will crash immediately. Therefore the paper proposes a self-adaptive strategy which does CPU scheduling in Xen. It spontaneously traces the change of CPU status and migrates virtual processors mapped in faulty CPUs to available CPUs. Experimental result shows that the virtual machine can keep running steadily and the performance of workloads in Xen reduces smoothly while intermittent occurs frequently.

Key words: Xen system, virtualization, multicore technology, intermittent fault, self-adaptive strategy

摘要： Xen虚拟化环境没有考虑CPU的间歇性故障带来的影响。基于此，建立模拟CPU间歇性故障的时间模型，在该模型下未修改的Xen系统中的虚拟机会立刻崩溃。提出一种自适应的策略来改进Xen的CPU调度，该策略主动跟踪CPU的状态变化，将发生故障的CPU上的虚拟处理器迁移到可用的其他CPU上。实验结果表明，当CPU间歇性故障频繁发生时，应用该策略可以使虚拟机继续稳定地工作，性能平滑地降低。

关键词: Xen系统, 虚拟化, 多核技术, 间歇性故障, 自适应策略

CLC Number: 

• TP391