Congestion Prediction: from Metric Definition to Routing Estimation

In this paper, an accurate congestion prediction algorithm is presented which is based on congestion metrics evaluation and router´s behavior. Various congestion-related metrics were defined and evaluated using one of the most recent routing estimation algorithms. Our experiments show that overflow metric results is more accurate estimation than others. In additions, the bend distributions after detailed routing for IBM-PLACE benchmarks were extracted and used to guide the probabilistic method. Furthermore, router´s behavior for congestion minimization was modeled and used to propose an accurate prediction method. The experimental results show that for IBM-PLACE benchmarks, our algorithm estimates the congestion more accurately than a recent method by about 19%.