Real Time Traffic Inference Model for BE Traffic QoS improvement using GT slack in NoC Systems

摘要

Network on Chip routers have become more sophisticated for various applications, where the quality of service also becomes more important due to the increased use of mobiles, PDA devices and many more. To increase the quality of service in Network on the Chip routers, we propose a Real Time Traffic Inference (RTTI) model with a little exchange of traffic patterns among the routers. The routers in the proposed model exchanges the traffic pattern without using a specific control message which reduces the additional communication overhead on the network. Each group of routers uses different timers and time slot at which it appends the traffic pattern occurred in the downlink. In the presence of traffic pattern, the router performs little validation on the packets arriving at each port and updates the traffic matrix consequently. Each router maintains a short term traffic cache, where it stores the traffic patterns of each neighboring routers. The router infers the traffic maybe present at a destination router before selecting a router to transfer the packet towards the destination.