Congestion metrics distributed over an area, instead of looking for a single congestion parameter such as wire length, are discussed in this paper. The authors present their case well, and provide a simple metric, as well as an approach to solving an important problem. They do this in the form of a simple algorithm, and conduct a number of experiments using benchmarks.
The authors conclude that the congestion correlation and its mapped net list are accurate enough for predicting congestion. Experimental results showed a 37 percent reduction, on average, in track overflows, as compared with conventional technology mapping. The authors conclude that they have introduced a distributed metric to predict routing congestion that showed a consistent improvement in the congestion and wiring distribution. I recommend this paper to all who are involved in routing. It is very well written, and easy to read.