负载类型相关的Xen虚拟机系统性能模型

针对Xen虚拟机系统执行网络I/O密集型负载时容易耗尽Domain0的CPU资源而过载和执行计算密集型负载时在客户域平均性能与数目之间存在线性规划的问题,提出了两个负载类型相关的性能模型。首先,通过分析Xen虚拟机系统处理网络I/O操作的CPU资源消耗规律,建立了CPU核共享和CPU核隔离两种情况下的客户域网络I/O操作请求次数计算模型；然后,通过分析多个相同客户域并行执行计算密集型负载的平均性能与一个相同客户域执行相同负载的性能表现之间的关系,建立了并行执行计算密集型负载的客户域平均性能分析模型。实验结果表明,两个性能模型能够有效地限制客户域提交的网络I/O操作请求次数以防止Xen虚拟机系统过载,并求解给定资源配置情况下执行计算密集型负载的Xen虚拟机系统客户域伸缩性数目。

Workload Type-dependent Xen Virtual Machine System Performance Models

In order to avoid the overload of Xen virtual machine system executing network I/O-intensive workload caused by the CPU resource exhaustion of Domain0 and solve the linear programming between the average perfor-mance and the number of guest domains when executing computing-intensive workload,two workload type-dependent perfor-mance models were proposed.Firstly,the network I/O request number computing models under the constraint of CPU core-sharing or CPU core-isolation were established by analyzing the CPU resource usage law of Xen virtual machine system executing network I/O-intensive workload.Secondly,the average performance analyzing models of guest domains executing computing-intensive workload were established by analyzing the relation between the average perfor-mance of multiple uniform guest domains concurrently executing computing-intensive workload and the performance of a same guest domain executing the same workload.The experimental results indicate that two performance models can effectively limit the network I/O request number of guest domains to prevent Xen virtual machine system from overloading and figure out the scalability number of guest domains in Xen virtual machine system executing computing-intensive workload with given computing resource respectively.