On the Cost of Shifting Event Processing within Wireless Environments

With the emergence of wireless sensor networks, the issues of event recognition and processing have been partially shifted into the embedded domain. New processing capabilities on small devices allow for physically close event monitoring and fast filtering without having to set up a wired infrastructure beforehand. This opportunity for flexible deployments, local data storage and demand-driven event forwarding opens up new application areas for event-centric architectures. However, the convenience of localized event processing comes at a cost, such as sparse resources or medium contention when relying on wireless communication. Several parameters have to be evaluated to decide whether pushing the application logic into a sensor network is worthwhile, or whether a conventional server-centered deployment is to be preferred. In this paper, we discuss parameters influencing an architectural decision and their interdependencies, illustrate our contribution with the help of an example and provide a generic cost model for estimating this decision.