Processing streams of events generated by sensors and apps has become key to the success of many important applications, ranging from social networks, through wearable medical sensors, to the industrial Internet. However, writing stream processing applications presents engineers with a real challenge. Difficulties include: handling events arriving at high rates; distributing analytics over a set of devices ranging from those at the edge (e.g. wearables, field gateways, sensors and phones) to servers (e.g. in a cloud); and, meeting non-functional requirements such as energy, networking, security and performance. Rather than place all the responsibility for meeting these challenges in the hands of the application programmers, we are exploring an alternative approach which overcomes these difficulties through tools that automatically optimise and deploy the stream processing code across sensors, field gateways and clouds. Enabling this approach requires that both the functional and non-functional requirements are described declaratively. The talk will describe the system and present experimental results for a healthcare (diabetes) application.
