Visualizing High-Performance Cluster Health, Performance, and Security

Research Intranet

Brief Description: A Visualization System to Monitor Cluster Health, Performance, and Security

Research Purpose: The essential problem to long-term cluster operation is that state-of-the-art cluster monitoring tools simply provide too much information for direct diagnosis. Every cluster is monitored continuously, with a number of measurements taken at regular intervals. Our proposal is to develop a cluster health analysis and visualization system which assimilates all of the information from existing health monitoring software, finds patterns in the various logs, and displays these patterns in an integrated visualization and query space. The system administrator will use the monitoring application to check the cluster’s work load and will make sure the applications are evenly distributed over the clusters in case any application is taking too many resources. This will enable the proper functioning of the clusters and will enhance the speed of the operations. The monitoring application will be web-based, thus enabling the system administrator to access the cluster information from any computer that has an internet connection.

Publications

Software

• In Progress

Research Team

• Mike Wong, David Frank, Dung Lai, Dragutin Petkovic