Service discovery and selection in IoT: A survey and a taxonomy

Recently, Internet has evolved into a new generation, called Internet of Things, thus enabling the connection between the physical and the digital worlds by creating an ubiquitous and self-organizing network. A huge number of smart objects are becoming now identifiable and addressable while being able to communicate with each other. Moreover, the integration of cloud infrastructures in the design of IoT, has moved this new trademark technology into a new dimension, enabling virtualisation and service provisioning. Billions of cloud services with different performance levels, requirements and functionalities are thus being offered in IoT, raising however the issues of their management, discovery and selection. In the literature, a considerable effort has been invested to address service discovery and selection in the context of IoT, despite the lack of standardization that meets the IoT requirements. In this paper, we propose an exhaustive taxonomy to classify service discovery approaches in the context of IoT, that we subsequently evaluate according to different aspects and criteria. Then, we discuss the gaps and advantages of each class of our taxonomy and locate the context and the requirements under which each can operate. Finally, we identify the challenges and future research directions in this domain. •We survey the most recent and interesting works in the literature that were dedicated to design SD and selection approaches in the IoT context. Then we discuss and compare their characteristics and their functioning.•We put forward in this paper a much exhaustive and elaborated taxonomy that includes several aspects to identify and classify IoT service discovery and selection approaches.•We conduct a comparative study of the different classes of solutions, considering several aspects and criteria. Then, we report the advantages, the limitations of the each class of our taxonomy.•We respond to some research questions and highlight guidelines to overcome the challenges of IoT service discovery and selection: What are the advantages and the limitations of using each class of methods relatively to each aspect of the SD system: architecture, description, and SD algorithms?; Which are the most dominant classes in the literature, and how the tendency is evolving in the last three years? ; Which descriptive methods, architectural designs and implementation models are the most considered to operate with each class of SD and selection algorithms? ; Which QoS met