When we designed OGC SensorThings API, efficiency and scalability are two critical requirements.  SensorThings API needs to be efficient so that it can accommodate battery power-constrained, computing power-constrained, and low cost devices. SensorThings API also needs to be scalable because it needs to scale to accommodate a very large number of IoT devices.

Achieving these two requirements provides the flexibility allowing us to innovate by configuring SensorThings with different network model. For example, SensorUp’s SensorThings solution can be deployed with a centralized model, a decentralized model or a hybrid model. A centralized model means there is a single point of entry (e.g., an api endpoint) provided to the clients/sensors. In a centralized model, a very powerful centralized server or clusters offer the SensorThings service for all clients/sensors. Today’s popular Software-as-a-Service (SaaS) model is an example of such configuration.

Screen Shot 2016-02-19 at 5.51.49 PM

However, some application requires a decentralized solution. The decentralized models means the various SensorThings server nodes work collaboratively as a single and coherent system of systems. Logically it performs as one system but physically the storage and computation are distributed in the local network. Take emergency management as an example. When a disaster strikes (e.g., earthquake), Wide Area Network (WAN), such as the Internet, very possibly will not be available. However, the rescue teams still need an IoT platform that integrate various sensing systems in order to compose a Common Operating Picture (COP). In this case, a decentralized SensorThings solution can be very useful. SensorUp’s solution allow different first responder groups to set up their own local SensorThings effortlessly. And these SensorThings can work collaboratively to form a single system of systems. When the Internet is available again, these local SensorThings nodes then provide a seamless transition and join the other SensorThings nodes on the Internet. In addition, such decentralized model is more robust as there is no single point of failure. It also offers better security and privacy, because there is no single point for network security breach.

At SensorUp we are working on a compact version of the SensorThings server, specifically for Raspberry Pi.   These Pi-based SensorThings can be deployed in the field on demand, and provide great flexibility and reliability. Stay tuned for more!

In fact, a decentralized architecture for IoT was my PhD thesis topic. Curious readers can read my paper titled “A New Peer-to-Peer-based Interoperable Spatial Sensor Web Architecture” for more details. Or please leaves comments here to let us know your thoughts.

Reference

Liang, S. A New Peer-to-Peer-based Interoperable Spatial Sensor Web Architecture. Proceedings of ISPRS Congress Beijing 2008, Beijing, China, July 2008; 37, pp. 1009–1014.