A Dynamic Declarative Composition Scheme for Stream Data Services

Abstract

With the fast development of Sensor Network, Internet of Things, mobile devices, and pervasive computing, enormous amounts of sensor devices are deployed in physical world. Data streams produced by these sensor devices, deployed broadly, can be used to create various value-added applications. Facing continuous, real-time, high-frequency, low-valued data streams, how to flexibly and efficiently cooperate them for creating valuable application is very crucial. In this study, we propose a service-oriented manner to realize flexible streams integration. It considers data stream produced by one sensor data as a stream data service and utilizes composing multiple services to realize the cooperation among sensor devices. Firstly, we propose a stream data service model based on Event-Condition-Action rules, which can encapsulate steam data as services and continuously and timely process stream data into value-added events. Then, we propose a declarative method which can dynamically compose stream data services. Based on two kinds of declarative rules, that is, sink-rules and connect-rules, multiple data streams can be dynamically integrated through flexible service composition. To ensure the performance of service composition, we also employ a sensor partition strategy and process multiple service compositions in parallel. Through comprehensive evaluations by experiments, our service composition method shows both good efficiency and effectiveness.