Title: HiperJobViz: Visualizing Resource Allocations in High-Performance Computing Center via Multivariate Health Metrics

Abstract—Scheduling, visualizing, and balancing resource allocations in High-Performance Computing Centers are complicated tasks due to a large amount of data and the dynamic natures of the job scheduling and resource allocation problem. This paper introduces HiperJobViz, a visual analytic tool for visualizing the resource allocations of data centers for jobs, users, and resource usage statistics. The goals of this tool are: 1) to provide an overview of the current resource usages, 2) to track changes of resource usages by users, jobs, and hosts, and 3) to provide a detailed view of the resource usage via multi-dimensional representation of health metrics, such as CPU temperatures, memory usage, and power consumption. To support these goals, our visual analytics tool provides a full range of interactive features, including details on demands, brushing and links, filtering, and ordering. The visualization tool is demonstrated on the HPC center of 467 computing nodes.

Index Terms—Multivariate analysis, Parallel Coordinates, Customizable Radar Charts, Job Scheduling, Health Metrics, Power Consumption, High-Performance Computing Centers