T4.2 Intelligent Reasoner for Iterative and Adaptive Analysis

Lead: FTW

The second and most important component of mPlane Supervisor is the intelligent reasoner which is able to correlate and reason on the results provided by the analysis modules on T4.1 to get more accurate and detailed insights, additionally triggering new actions to improve analysis and drilling down particular issues.

These actions may include collecting additional data at different granularity at WP2, analysing different types of data from WP3, using data from additional probes, or even dynamically instantiating new measurement functionalities on them. The reasoner provides the necessary dynamic adaptation to the intelligent monitoring system.

Such flexibility in data collection and analysis additionally permits mPlane to adapt to the varying system condition and requirements. For example, some type of measurements initially used for the analysis may be changed to other type of measurements which provide the same insights with less overhead and better efficiency (i.e., energy consumption, bandwidth utilization, etc.)

The intelligent reasoner has access to a set of domain-knowledge-based rules that helps him to correlate intermediate analysis results with new monitoring and analysis actions in the iterative analysis process. These rules are conceived as a knowledge structure where mPlane users can include domain-knowledge information related to the different monitoring applications that enrich the reasoner analysis. The set of rules is not necessarily static or only adapted by domain-knowledge information, but can eventually be expanded by learning from past experiences.

This knowledge structure will provide certain functionalities in the form of APIs that will allow the user to define particular knowledge-domain-based decision rules and to execute particular tasks depending on the intermediate results of the analysis, always considering the particular use cases addressed by mPlane.