Design And Realization Of Open Stack-oriented Real-time Monitoring And Analysis System

Monitoring is vital for cloud computing platform's high availability and stability.A more dimensional monitoring scheme is needed by the platform because of its complex environment.The elastic demand of users for cloud computing and storage also requires monitoring framework to have a higher ability to a nalysis data in real time.In conclusion,the challenges of nowadays cloud monitoring technology are: the need of a comprehensive and fine grit monitoring method,and the analysis ability of real-time monitoring data.Currently,existing monitoring schemes can't get virtual machine's internal monitoring data.They also haven't established effective model of monitoring and analysis.And processing flow of monitoring data can't satisfy the requirement of real-time analysis.Monitoring schemes remain to be perfect.In the light of the problems above,this paper deeply researched existing technology of monitoring and complex event processing,proposed a new solution of monitoring,the main contributions of this paper are:(1)We do the research of Open Stack project's workflow and its components' communicating mechanism,study the monitoring framework Ceilometer,and focus on ceilometer's acquisition method of monitoring data.On the basis of source code research,we write plugins to acquire the monitoring data of virtual machines,and modify the setting file of Ceilometer framework to change the destination of monitoring data flow.(2)We establish a three-dimensional model oriented to the cloud computing platform.This three-dimensional model contains full a mount of platform's monitoring metrics: its horizontal axis represents the physical machines,longitude axis represents the applications and vertical axis has physical section,virtual section,application section and business section to represent the different kinds of monitoring metrics.This model can identify and choose monitoring metrics more precisely and rapidly.Furthermore,it can differentiate metrics of same application on different physical machines and metrics of different applications on same physical machine.This model can reduce the negative influence caused by metric confusion.(3)We use the engine based on complex event process to make the function of data's real-time analyzing come true.We build rules in the esper engine,and epser will process real-time monitoring data according to the rules.After that monitoring scenario will be recognized by esper.(4)We design the main module of the solution mentioned above.O n monitoring data collection layer,we bring a monitoring model of cloud computing and choose context of analysis based on given scenario;on real time analysis layer,we design the analyze rule for monitoring data;on scenario response layer,we design the management instruction set.We complete the solution based on Open Stack cloud computing platform and combine Ceilometer and CEP engine.Finally,we test and verify the feasibility of this solution.