Traffic forecaster for MPLS multimedia data streams
Barry Ambrose, Freddie S. Lin · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2006
Contemporary high performance data networks carry a wide range of multimedia services (voice, video, audio, text, sensor data, etc.) that require an outstanding Quality of Service (QoS) to provide performance guarantees in priority delivery, latency, bandwidth utilization, load balancing, etc. With the advent of recent Multi-Protocol Label Switching (MPLS) network standards, the QoS has made significant progress in performance to provide these performance guarantees. Right now, attention has turned to the task of managing these QoS networks more efficiently through the handling of network traffic. We have investigated a novel Network Traffic Forecasting Assisted QoS Planner technology that will provide constantly updated forecasts of data traffic and server loads to any application that needs this information. Using source models of voice, video and data traffic based on empirical studies of TCP/IP data traffic carried out by Paxson and Floyd in the 1990's, our studies have shown that the system may provide up to 43% bandwidth savings for MPLS data streams, by predicting future traffic flows and reserving network resources to accommodate the predicted traffic. The system additionally provides a means to submit bandwidth reservation requests for those applications that need assured service guarantees for data delivery. The technology essentially increases the efficiency and effectiveness of multimedia information and communication network infrastructure that supports multiple or adaptive QoS levels for multimedia data networking and information system applications.