Research Of The Real Time Anomaly Detection And Auto Reaction System Based On Cloud Computing

Recently, cloud computing has been widely implemented by manycompanies and individuals for its flexibility, convenience and affordablemeans to access storage and other services. The technique itself hasbecome mature as well. With cloud computing becoming more and moremature, safety and consistency are viewed as the fatal problem. Therefore,a lot of works on anomaly detection in cloud computing based onperformance statistics have been put forward and implemented. Meanwhile,virtual machine migration serves as an important virtualization techniquehas become sophisticated nowadays. However, virtual machine migrationmakes performance statistics from the new hosts inconsistent with the old,which invalidates those existing detection methods. In this work, ananomaly detection and response system is proposed to not only performanomaly detection with the scenario of environmental changes, but alsopinpoint possible reasons for detected anomalies and response to them.Thus, the time interval between anomaly takes place to it is caught will bedecreased. In this way, the damage to cloud platform will be minimized.In this work, an adaptive anomaly detection and reasoning algorithmis raised based on the context of environment variation in cloud computing.It is based on cloud performance statistical data mining and gives real-timeanomaly alarms and reason reports. This algorithm has good adaptabilityas it maintains accuracy even cloud computing environment changes.Experiments indicate that the proposed algorithm has both better anomalydetection rate and lower false alarm rate than other classic anomalydetection algorithms. At the same time, it has great real-time performance and consistency. As the aspect of reasoning performance, experimentspoint out that this algorithm is able to successfully provide the reasons ofanomalies.Besides, this work also proposes an anomaly response mechanismbased on the category of anomalies. This work gives optimized responsemeasures and degree of response to keep the security of cloud computingfurther with the help of Analytic Hierarchy Process. The workload ofadministrators· is therefore mitigated and the speed to exclude anomalies isaccelerated.