Distributed Monitoring System Based On Complex Event Processing

As an indispensable component of large-scale distributed systems, the monitoringsystem can help relevant people to have a better understanding of the runtime status ofthe distributed system, to observe and prevent system failures, and to react timely toanomalous situations.This thesis proposed a event-driven approach for monitoring distributed systems.It is a efcient, intelligent and extensible distributed monitoring system. It collectsmonitoring data in a fexible way using an extensible and dynamically confgurableprobe platform, employs a publish/subscribe network for distributing data to automat-ically detect newly-joined resources and to avoid transporting the non-demanded data,processes massive monitoring data in a real-time manner with the Complex Event Pro-cessing (CEP) technology to discover non-trivial events, and uses an intelligent migra-tion policy for CEP rules in the purpose of making CEP distributed and saves networkbandwidth.TheComplexEventProcessing(CEP)isanevent-drivenreal-timedataprocessingtechnology. TheCEPenginecananalysesimpleeventsthatenterstheengineaccordingto defned event processing rules, and produces complex events when some rule ismatchedbyoneorasetofevents. Itcananalyseandcorrelatemultipleeventstoextractinteresting situations from massive trivial events and thus make the system monitoringwork more efcient. Meanwhile, CEP has advantage over traditional DBMSs in thatCEP ofers the capability of continuous query and real-time processing, and is suitableto be employed into monitoring systems that have the requirement for processing ina timely manner. Since currently some relatively complex monitoring requirementscannot be expressed in available CEP engines, we decided to implement a novel Java-based extensible CEP engine, which provides a set of APIs for users to defne custom event processing rules in Java.The thesis introduces the proposed distributed monitoring system in aspects ofarchitecture, design and implementation, and provides a case study on the deploymentof the system. The design, implementation and the extensible API of our CEP engineis introduced; practical examples of complex event processing rules are provided toprove the expressiveness and extensiveness of our engine. In addition, we carried outperformance benchmarking of the CEP engine.