Abstract

Applying IoT technologies to cities and buildings can not only provide intelligent services to the users but also better utilize resources. While developing applications for these two domains, there are several challenges. Manageability, connectivity, and programmability are three of the major challenges. This paper presents the design and study of programming applications for smart cities and smart buildings. Data-driven programming model is designed to simplify the complexity on programming applications on large scale devices. To reduce the complexity of inter-connecting large numbers of devices, three message exchange models supported at WuKong, an intelligent virtual middleware for IoT applications, are studied in this paper: cloud-based, NDN-based, and peer-to-peer message exchange models. Cloud-based model is the most intuitive model but is only suitable for timeinsensitive applications. The other two models can support time sensitive applications: NDN-based model can support large scale deployment but leads to higher hardware cost; peer-to-peer model can be employed to the applications in the scale of building without heavy hardware cost. Our evaluation results show that cloud-based model performs worse than the other two models when the number of senders/receivers in the order of magnitude. Last but not the least, security is the foundation of the smart city/building applications. CapeVM, the core of WuKong virtual middleware, takes the advantage of ahead-of-time complier to prevent illeagal memory access and execution order without excessive performance overhead.