VOtus : A Flexible And Scalable Monitoring Framework for Virtualized Clusters
暂无分享,去创建一个
Large-scale distributed processing frameworks such as Hadoop are currently enjoying wide popularity for big data computation. Performance Analysis and monitoring under these frameworks are inherently difficult especially in a virtualized environment. Existing distributed monitoring tools can only report virtual resource usage. Such reported information might be insufficient for developers and users to acquire deep insights into the performance of distributed applications. This paper describes VOtus, an initial step towards extending the Otus monitoring tool for virtualized clusters on private clouds. By collecting supplementary metrics from the hypervisor (when available), VOtus allows users to effectively and efficiently monitor a virtualized cluster. It also provides enhanced comprehension of distributed applications, which helps in answering performance related queries that relate to capacity planning, placement, and migration of virtual machines on the cluster.
[1] David E. Culler,et al. The ganglia distributed monitoring system: design, implementation, and experience , 2004, Parallel Comput..
[2] Randy H. Katz,et al. Chukwa: A System for Reliable Large-Scale Log Collection , 2010, LISA.
[3] Garth Gibson,et al. Otus: resource attribution in data-intensive clusters , 2011, MapReduce '11.