Study of active subscription control parameters in large-scale smart spaces

摘要

The development of smart spaces for Internet of Things (IoT) environments meets the scalability challenge since many participants are involved into the distributed computation. In particular, many sensors continuously provide data, many reasoners analyze the sensed data to construct services, and many mobile clients regularly join and leave the smart space to consume the services. The network interaction is information- driven, using a semantic information broker, which implements a passive strategy for subscription. The strategy leads to performance bottleneck. We continue our study of the active control strategy, when a part of subscription processing is delegated to the subscribers. The client multiplicatively decreases its check interval, if subscription notifications are lost, and additively increases the interval, otherwise. We study the role of strategy parameters: the client can select their values preliminary and keep fixed, or the client can change them dynamically depending on the recent situation. With the aim we provide mathematical model which represents important performance metrics as a function of strategy parameters. Analytical result are validated by experimental evaluation. Additionally, our simulation experiments evaluate the scalability in dependence on the number of participants.