Towards a Generic Architecture for Interactive Cost-Aware Visualization of Monitoring Data in Distributed Systems

Abstract

The collection of monitoring data in distributed systems can serve many different purposes, such as system status monitoring, performance evaluation, and optimization. There are many well-established approaches for data collection and visualization in these areas. For objectives such as debugging complex distributed applications, in-depth analysis of malicious attacks, and forensic investigations, the joint analysis and visualization of a large variety of data gathered at different layers of the system is of great value. The utilization of heavy-weight monitoring techniques requires a cost-aware on-demand activation of such monitoring. We present an architecture for an interactive and cost-aware visualization of monitoring data combined from multiple sources in distributed systems. We introduce two distinguishing properties: the possibilities to reconfigure data collection and a cost prediction mechanism that supports the user in a cost-aware, dynamic activation of monitoring components in an interactive in-depth analysis. We illustrate the use of such cost prediction for monitoring using VMI-based mechanisms.