MS.17 Semantic metadata processing for system monitoring

Metadata are the main resource for understanding of the user
interaction with the system. System logs where different events are
recorded, explicit user feedback about the application being used,
and similar metadata give insights about the user satisfaction with
the system and indicate possible system improvements. Moreover, these
metadata can also be processed to detect implicit events and provide
assistance for users if certain issues in the system are detected.
Different patterns in system logs can indicate that there is an issue
that needs to be addressed. Moreover, these patterns can have
different levels of urgency depending on different factors, e.g., the
frequency of the pattern occurrence or metadata types included in the
pattern.

To benefit the most from these metadata and support their sharing and reuse, the metadata should be represented with a semantic-aware formalism such as RDF. Furthermore, such representation also supports their exploration in a systematic manner, e.g., via SPARQL - the standard query language for RDF. Once modeled and populated, these
metadata can be processed in real-time to raise the alerts about
possible events happening in the system.

This topic will focus on the modeling and processing of metadata for
system monitoring. The metadata model should capture a set of quality
attributes that are monitored for a system and represent these
attributes with a semantic-aware formalism. Related to their
processing, the emphasis will be on the definition of the ECA
(event-control-action) rules that will be used to raise the alerts
when needed. As quality attributes and ECA rules are directly related,
the topic will explore the relation between the two and propose
techniques and/or methods on how to design optimal solutions in this context. Additionally, this topic will deal with more complex
scenarios that involve other kinds of metadata such as user feedback,
and propose how to process them together with the existing quality
attributes coming from the system monitoring. Moreover, the
combination of different ECA rules can produce implied/derived ECA
rules and this possibility will also be considered.

Main Advisor at Universitat Polit├Ęcnica de Catalunya (UPC)
Co-advisor at TBD