Joule: A Real-Time Framework for Decentralized Sensor Networks

Traditional Internet of Things (IoT) sensor nodes transmit raw data to centralized servers for analysis and storage. At higher data rates this architecture is no longer feasible. Instead data must be processed at the lowest level possible to include the sensor node itself. The Joule framework enables decentralized data processing by distributing computation into modules which are dynamically linked at run time by a process supervisor. Modules may be collocated on a single machine or distributed across a network. Data is stored where it is processed creating a fully decentralized architecture. Processed data can be queried by centralized servers or delivered directly to the end user. This allows high bandwidth sensors to be incorporated seamlessly into existing IoT deployments without depleting network bandwidth or server resources. This paper presents the Joule framework as well as performance benchmarks and deployment case studies.