Abstract: The cloud platforms aggregate a lot of computing resources and adopt Docker technology to achieve dynamic resource scaling and provide services on demand. Docker, a kind of lightweight virtualization technology which has the characteristics of resource isolation, rapid deployment and low cost, is widely used in the construction of the cloud services so that Docker-based containers has become the important basis of core cloud businesses. Therefore, in order to manage the large-scale cloud cluster and enforce the quality of cloud services for consumers, Monitoring mechanism for the container-based clouds are indispensable. In this paper, we design and implement a cloud monitoring system - PLMSys based on cluster performance and container logs. It provides the following functions: (1) Multi-dimensional resources monitoring. PLMSys can monitor the running state of the cluster host and container, including the utilization of CPU, memory, disk and other resources. (2) Container log collection.PLMSys can centrally collect the logs generated by all containers of the cluster. (3) Rule-based exception alerts. PLMSys allows users to define the abnormal state of the host and container by creating rules, and provides multiple alerting methods. (4) Workload analysis and prediction. PLMSys extracts the descriptive statistics from the cluster workload and uses the time series models to predict the future workload. (5) Data monitoring visualization. The system uses rich visual charts to reflect the running state of cluster host and container.By using PLMSys, users can better manage cluster hosts and containers.
Authors: Yongzhong Sun (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China); Kejiang Ye (Chinese Academy of Sciences, China); Cheng-Zhong Xu (University of Macau, China)
Email: syunwingzung@foxmail.com, yekejiang@gmail.com, czxu@um.edu.mo