A practical approach for enabling online analysis of event streams
Sebastian Salvucci, Mariano A. Cilia, Alejandro Buchmann · 2007
The event-based paradigm has gained interest as a solution for integrating large-scale distributed and loosely coupled systems. It also has been propelled by the growth of event-based applications like ambient intelligence, utility monitoring, event-driven supply chain management (eSCM), multi-player games, etc. The pub/sub paradigm is particularly relevant for implementing these types of applications. The core of a pub/sub system is the notification service and there are several commercial products, open source and research projects that implement it. The nature of notification services (distribution, scale, large amount of concurrent messages processed) as well as the diversity of their implementations, make their run time analysis difficult. The few existing solutions that address this problem are mainly proprietary and focus on single aspects of the behavior. A general solution would significantly support research, development and tuning of these services. This paper introduces a notification service-independent analysis framework that enables online analysis of their behavior based on streamed observations, flexibly defined metrics and visual representations thereof.