engleski

Performance Modeling of Publish/Subscribe Middleware Components

Internet of Things has changed the environment dramatically, it introduced the all-time connectivity at the level of individual user and enabled monitoring of all user activities by connecting various sensors and data sources into the IoT ecosystem. This was supported by publish/subscribe infrastructure that efficiently interconnects all data flows. In such, global, dynamic and heterogeneous environment satisfactory performance of the underling layer for information exchange is crucial in order to support IoT services in production deployment. Therefore, to plan and support all services it is necessary to investigate the performance of the publish/subscribe middleware in details, as opposed to the overall performance analysis where the publish/subscribe middleware is analyzed as black-box. Detailed analysis provides insights in the performance of the whole system, identifies possible bottlenecks and provides information about maximum throughputs that can be handled with given parameters. In this paper we extended the CUPUS middleware with the performance monitoring components that allowed detailed insight into processing time at internal components. This allowed us to model processing times at individual component with well-know distribution (such as exponential). Such data can be used to perform complex analysis of performance of the whole middleware.

publish/subscribe middleware, performance, modeling, processing time

nije evidentirano