Workflow Support for Low-Power Wireless Sensor and Actuator Networks

• Main Author: Guerrero, Pablo E.
• Format: Others
• Language: German; en
• Published: tuprints 2014
• Online Access: https://tuprints.ulb.tu-darmstadt.de/4283/1/phd-thesis-guerrero-tuprints.pdf; Guerrero, Pablo E. <http://tuprints.ulb.tu-darmstadt.de/view/person/Guerrero=3APablo_E=2E=3A=3A.html> (2014): Workflow Support for Low-Power Wireless Sensor and Actuator Networks.Darmstadt, tuprints, Technische Universität, [Ph.D. Thesis]

A decade ago, the advances in the miniaturization of electronic components made it possible to integrate three fundamental functions into a tiny, battery-operated device, namely sensing, processing and wireless communication. This gave rise to a new family of computers that, when connected with each other in masses, are known as low-power, wireless sensor and actuator networks (WSANs). These networks are the enabling technology of the Internet of Things, a market that is predicted to encompass ∼14 billion devices by 2020. As the initial challenges of the technology are overcome, such as identifying adequate medium access control protocols, localization techniques, and network standards, the range of possible applications has grown up. To date, however, there is general consensus that the existing node-level programming languages do not provide adequate abstractions to implement user applications. Indeed, the predominant approach is very close to the hardware and involves the C programming language (or a variant of it). This makes it difficult for domain experts to employ the technology without a sensor network specialist. The research community has proposed a number of middleware approaches to simplify the development effort. However, these macroprogramming languages mainly focus on data extraction, and not on in-network actuation. In this work we advocate the usage of workflows as a means to define the logic that orchestrates the network activity. With this abstraction, the loop of event-sensing, decision and actuation can be closed, leading to a reduced need for slow and error-prone human intervention in the process. In this way, the whole WSAN loop can be shifted to the network. Our main contribution consists of the conception of a holistic workflow modeling and execution platform for WSANs, together with the design and implementation of ukuFlow, a workflow platform for low-power nodes, that runs entirely in-network, not requiring an external server infrastructure. Secondly, we present the ongoing work on the development of TUDμNet, a metropolitan-scale federation of sensor network testbeds, with which the empirical evaluation of ukuFlow, among many other research projects, was carried out. We have identified a set of operators to compose workflows, and provided a lightweight architecture that controls the flow of such processes with an in-network execution algorithm. We present a detailed evaluation of various performance indicators for each major component of the architecture, including the data manager, command execution, and complex event detection modules. The results of the evaluation show the feasibility of the approach, in spite of the adverse resource constraints and the tough network settings employed. We strongly believe that this abstraction is of practical relevance to WSAN practitioners, while still holding promise to an in-network operation.