Cognitive Visual Analytics of Multi-Dimensional Cloud System Monitoring Data

Contributions

In this work, we present a new design for interactive cognitive visualization to provide effective analytical capability of intelligent cloud systems. For each target, the data item composed of multiple parameters being monitored can be regarded as a point in a multi-dimensional space. The data items are collected over time at specific frequency commensurable with computing tasks, i.e. milliseconds.

Our contributions are as follows:

• ●

  We put forward a framework to combine components in cloud monitoring and data visualization seamlessly. The visualization design aims at showing data patterns that users and administrators are mostly concerned about, for instance, underutilization or overutilization of resources, e.g. memory, disk, processors

• ●

  By aggregation and clustering of raw monitoring data with new analytic approaches and reference vector settings, we aggregate monitoring targets into three states: cool, norm and warm. This is explained in section of Methods.

• ●

  In order to adapt to the elastic features of a cloud system, we design a Level-of-Detail (LOD) visualization approach, and present its advantage with case studies.