Internet of Things (IoT) systems produce greatudamount of data, but usually have insufficient resources toudprocess them in the edge. Several time-critical IoT scenariosudhave emerged and created a challenge of supporting low latencyudapplications. At the same time cloud computing became a successudin delivering computing as a service at affordable price with greatudscalability and high reliability. We propose an intelligent resourceudallocation system that optimally selects the important IoT dataudstreams to transfer to the cloud for processing. The optimizationudruns on utility functions computed by predictor algorithms thatudforecast future events with some probabilistic confidence basedudon a dynamically recalculated data model. We investigate ways ofudreducing specifically the upload bandwidth of IoT video streamsudand propose techniques to compute the corresponding utilityudfunctions. We built a prototype for a smart squash court andudsimulated multiple courts to measure the efficiency of dynamicudallocation of network and cloud resources for event detectionudduring squash games. By continuously adapting to the observedudsystem state and maximizing the expected quality of detectionudwithin the resource constraints our system can save up to 70%udof the resources compared to the naive solution.
展开▼
