Traffic congestion is a compound result of multiple environmental and behavioral factors. Environmental factors are objective, refer to external conditions such as economic and population sizes, road conditions, resource locations and others. Behavioral factors are subjective, refer mainly to the routing behavior of travelers. Undoubtedly, travelers are selfish. They care only their own interests, and would like to follow paths minimizing their own travel times. Such selfish routing behavior will lead the system to an steady state in the long run. The state is usually a pure Nash equilibrium. This raises kind of an interesting question in terms of that whether such Nash equilibrium is harmful to congestion management. We devote this talk to answer the question. We shall prove that selfish routing is actually very efficient when total travel demand is large. So, congestion may not attribute to the selfish routing behavior of travelers. Moreover, we shall prove that resource locations are crucial to congestion management when the road network is crowded. To obtain these results, we shall introduce several new notions such as limit games and completely tightness, and a new technique called asymptotic decomposition of games.