| Monitoring and testing the virtual machine state accurately and timely is one of the important ways to optimize the cloud IaaS resource scheduling to improve the virtual machine migration mechanism. However, the forecast of virtual machine is often ignored in existing monitoring technology and the harm will be enlarged if the anomaly state of virtual machine monitor problem cannot be effectively solved. Therefore, the virtual machine state monitoring and anomaly detection technology research in cloud environments has important application value for improving the level of operation and management of all kinds of cloud platform.A method based on multi-attribute information entropy is proposed to detect anomalous states of virtual machines in cloud computing environments.Firstly, the 2-nonn of various attributes characterizing the states of a virtual machine is computed for each sampled data. Then, the joint information entropy is computed based on the occurrence frequency of each 2-norm value in a fixed period of time. Once the entropy reaches its maximum, the anomaly detection function in the proposed method is activated.During the anomaly detection, the moving weighted average and the variance of the 2-norm sequence are used to construct the test variable, and the non-parametric CUSUM algorithm is adopted to complete the anomaly detection. Experimental results based on Hadoop show the method cannot only reduce false alarms caused by accidental and transient anomalous states, but also give an accurate alarm before a significant anomalous state occurs.Finally, a multi-attribute abnormal locating method is proposed based on the characteristics of the anomaly detection method in this thesis. To reduce the time of finding out the problem for administrator, the abnormal location module will be activated once the method detect the abnormal state of the virtual machine and send out the alarm. It is concluded that monitoring location map of the attributes of the virtual machine can help the administrator find out the cause of the exception specifically and find out the corresponding solution. |
