Rest-based Internet of Things-Services and applications in a semantics-aware enterprise context

The Internet of Things (IoT) is the vision of a global infrastructure of networked physical objects. In order to use IoT in a semantics-aware enterprise there are some challenges: IoT-devices should integrate into enterprises as seamlessly as possible, both at a modeling level and at a technology level. At the modeling level, enterprise systems are often customized by non-software experts. In addition, more and more systems are applying machine-learning technologies. Both human modeling and machine reasoning need a precise semantic description of the entities they work with and their meaning. At the technology level, enterprise systems traditionally use a different protocol stack than IoT-devices. Neither these traditional enterprise protocols nor the IoT-protocols are semantics-aware. Furthermore, IoT-devices have properties that are typically unknown to enterprise information systems, such as their limited energy and consequently their maintenance needs (i.e. swapping batteries). Keeping this total cost of ownership low is one of the primary goals of IoT-operators. This thesis contributes to enterprise integration and semantic-aware integration by developing and evaluating two different approaches: a top-down approach and a bottom-up approach. The top-down approach scales down the existing OData enterprise protocol to very constrained IoT-devices. The bottom-up approach semantically enriches existing protocols. It consists of a service description language called Linked USDL for IoT, which allows to semantically describe services from a very high abstraction level, down to their technical realization. In order to utilize those approaches in an enterprise, we propose an architecture and abstractions that enable the integration of BPMN tools, semantic services, and constrained IoT-devices. We also investigated some reasons for the reluctance of developers to apply semantics, a behavior coined semaphobia in previous research. We evaluated our approach with an architecture evaluation method and through several experiments. The experiments were done on an experimental platform called Mote Runner on the following two hardware platforms: MEMSIC IRIS and Waspmote Pro. We propose an application-layer based framework for reducing the energy consumption by putting nodes to sleep (sleepy nodes). Sleepy Nodes were implemented and evaluated on the Mote Runner platform. We introduce a monitoring framework based upon high-level information that utilizes sleepy nodes to increase the network lifetime. As part of the monitoring framework, we present three different scheduling strategies: A simple first fit, an exhaustive strategy, and a strategy called dynamic partitioning. Dynamic partitioning is based on the observation that it is possible to combine measurements under certain circumstances. We were able to show that dynamic partitioning performs significantly better than first fit and only slightly worse than the exhaustive strategy