RES: Real-Time Video Stream Analytics Using Edge Enhanced Clouds

With increasing availability and use of Internet Things (IoT) devices such as sensors video cameras, large amounts streaming data is now being produced at high velocity. Applications which require low latency response surveillance, augmented reality autonomous vehicles demand a swift efficient analysis this data. Existing approaches employ cloud infrastructure to store perform machine learning-based analytics on This centralized approach has limited ability support real-time large-scale due network bandwidth constraints between source cloud. We propose RealEdgeStream (RES) an edge enhanced stream system for large-scale, performance analytics. The proposed investigates the problem by proposing (i) filtration (ii) identification phases. phase reduces amount filtering low-value objects using configurable rules. uses deep learning inference streams interest. phases consist stages are mapped onto available in-transit resources placement algorithm satisfy Quality Service (QoS) identified user. demonstrate that 10K element streams, with frame rate 15–100 per second, job completion in takes 49 percent less time saves 99 compared cloud-only based approach.