SPOTLIGHT Principles for Engineering IoT Cloud Systems

ecently, we’ve seen a wide adoption and deployment of Internet of Things (IoT) infrastructures and systems for various crucial applications,1 such as logistics, smart cities,2 and healthcare. This has led to high demands on data storage, processing, and management services in cloud-based datacenters, engendering strong integration needs between IoT and cloud services. Cloud services are mature and provide excellent elastic computation and data management capabilities for IoT. In addition, as IoT systems become complex, cloud management techniques are increasingly employed to manage IoT components. Thus, cloud services now act as computational and data processing platforms as well as management platforms for IoT. From a high-level view, IoT appears to be well-integrated with cloud datacenters to establish a uniform infrastructure for IoT cloud applications. However, the software layers on top of such integrated infrastructures are still fragmented, and therefore far from a uniform software layer to support a coherent execution environment for complex applications. IoT elements (sensors, actuators, gateways, lightweight applications, and so on) are developed, deployed, and operated separately from cloud services (such as storage and data processing). Because of the complexity of software ecosystems, IoT providers are increasingly different from cloud providers with regard to communication protocols, software layers, and provisioning models, to name just a few. In addition, although cloud services, such as load balancers, message-oriented middleware, NoSQL storage, and streaming data processing frameworks, are designed to accept workloads and data from IoT, they lack capabilities to be coordinated with IoT operations. For example, most cloud services reactively monitor the load from IoT and adjust their performance behavior, but rarely communicate back to the IoT elements to steer the load generated by the IoT. There are various reasons for these issues. Cloud services and IoT are created separately by cloud providers and IoT providers. In addition, the complexity of the IoT and cloud ecosystems prevents a single stakeholder from offering software that works well in IoT cloud systems. These issues prevent us from developing and operating IoT cloud systems in a co-