The IoT is gaining increasing attention. The overall aim is to interconnect the physical with the digital world. Therefore, the physical world is measured by sensors and translated into processible data, and data has to be translated into commands to be executed by actuators. Due to the growing interest in IoT, the number of platforms designed to support IoT has risen considerably. As a result of different approaches, standards, and use cases, there is a wide variety and heterogeneity of IoT platforms. This leads to difficulties in comprehending, selecting, and using appropriate platforms. In this work, we tackle these issues by conducting a detailed analysis of several state-of-the-art IoT platforms in order to foster the understanding of the (i) underlying concepts, (ii) similarities, and (iii) differences between them. We show that the various components of the different platforms can be mapped to an abstract reference architecture, and analyze the effectiveness of this mapping.