Continuous queries over distributed streams of heterogeneous monitoring data in cloud datacenters

The use of stream processing for state monitoring of distributed infrastructures has been advocated by some in order to overcome the issues of traditional monitoring solutions when tasked with complex continuous queries. However, in the domain of behavior monitoring the situation gets more complicated. It is mainly because of the low-quality source of behavior-related monitoring information (natural language computer logs). Existing approaches prevalently rely on indexing and real-time data-mining of the behavior-related data rather than on using event/stream processing techniques and the many corresponding benefits. The goal of this paper is to present a general notion of Distributed Event-Driven Monitoring Architecture that will enable an easy definition of expressive continuous queries over many distributed and heterogeneous streams of behavior-related (and state-related) monitoring data.