Search for: All records

Towards IoT-enabled world, the response time starting from data occurrence at the source until processed data delivery to the actuator is another QoS metric to be concerned. We call this requirement deadline. In coarse-grained stream processing, we could partially drop data in the streams with a specific drop rate to meet the deadline. This paper proposes an autonomic sampling method to decide the drop rate aiming at response time reduction oriented by the user-specified deadline. With consideration of processing and communicating time sharing among distributed worker nodes, we calculate a sampling number to satisfy the deadline requirement while preserving the maximum drop rate. The device will set a goal to maintain this sampling number for the next operating window. To evaluate the performance, we have implemented the proposed method on top of our previously-proposed stream processing engine called EdgeCEP. The results present that our proposed method can reduce almost 2-times latency and preserve a higher amount of request outputs compared to the fixed rate approach.