The performance analysis of Random Linear Network Codes is important both theoretically and for its applications. In this paper, we derive improved upper bounds for the failure probability of random linear network codes and analyze the limiting behavior as the field size goes to infinity. Unlike the previously reported bounds, the new bound is shown to be tight as the field size |F| goes to infinity.