A New Framework for Network Flow Queuing Delay Prediction Based on Stream Computing

In the network, accurately predicting network flow queuing delay is important for congestion control, network bandwidth allocation, and network performance improvement. In order to improve QoS (quality of service) and predict fine-grained queuing delay in advance, A framework is proposed in this paper. We firstly perform real-time preprocessing on fine-grained data packet, which is acquired using In-band Network Telemetry (INT) technology, based on the Storm platform. Then we use Spark Streaming to perform online prediction based on LSTM short-term prediction model which is obtained by training off-line data on the Spark platform. The online prediction and off-line training technology framework proposed in this paper will provide effective information for congestion avoidance or dynamic routing planning in real time.