T4.1 Monitoring Data Analysis Modules

Lead: ULG

The first component of mPlane Supervisor consists on a set of analysis modules. This task consists of the design and implementation of the analysis algorithms and modules, which are used to dig into the data gathered and pre-processed across the entire mPlane (at both W2 and WP3). The algorithms have a clear spatio-temporal notion of the problem they address, as they are capable of merging on the fly the local real-time analysis done by the probes designed by WP2 with the historical global data analysis performed by the technology devised by WP3.

The design of the different analysis algorithms will be restricted to the different monitoring applications supported by mPlane (decided from the use cases defined in WP1), but whenever possible, re-utilisation of the different algorithmic techniques will be envisaged. In this sense, analysis algorithms will be designed and implemented as analysis modules, which will expose different functionalities to its final user, i.e., the reasoner.

The complexity of the algorithms designed and implemented in T4.1 depends on the particular monitoring application which they are intended to serve, but in general terms, T4.1 algorithms are more specialized and have a broader picture of the problem they are tackling than in previous analysis stages performed either locally at WP2 probes or globally using WP3 techniques.

For example, let us consider the interaction between WP3-type data analysis and the analysis done by T4.1 algorithms; while WP3 performs basic statistical data analysis (e.g., number of packets per End-User per hour) on large amounts of historical and highly distributed data coming from the WP2 probes. T4.1 accesses only a reduced part of this pre-processed data for more elaborated analysis (e.g., analyse all those users who have sent more than N packets in the last hour, as offered by T3.3). Similarly, while probes at WP2 do local analysis at the vantage points where they operate, T4.1 algorithms combine and correlate the information coming from multiple of these probes and at different times to get more clear and global view of the analysis.

Data mining, machine learning, signal processing, and information theory approaches are the basis of T4.1 analysis modules. The algorithms designed and implemented at this task provide different degrees of insight based on the quality and type of the data they use. For example, limited insights if only coarser-grained, fewer dimensional data is available, and more accurate and detailed insights if richer, more relevant and finer-grained data is used. The insights provided by T4.1 are output to the intelligent reasoner designed in T4.2.