A Performance Analytical Approach Based on Queuing Model for Network-on-Chip
Zhang Youhui, Zheng Weimin, Dong Xiaoguo, Gan Siqing · 2010
To establish a highly-efficient analytical performance model of routers is crucial for the design of NoC. In this paper, an analytical router performance model which is based on M/D/1/B queuing theory is proposed to analyze various packet blockings at the input buffers during the transfer process, and then a computing method based on Markov chain for flow-control feedback probability is presented to estimate some key metrics in terms of buffer utilization, etc. Compared with BookSim, a famous cycle-accurate NoC simulator, the results show that the average error of the computing method for the flow-control feedback probability is 7.87%.