This paper presents a new mathematical model of AIMD (Additive Increase Multiplicative Decrease) TCP for general networks that we believe is better than those previously used when it is driven by bottleneck capacities. Extending the paper by Edmonds, Datta, and Dymond that solves the single bottleneck case, we view AIMD as a distributed scheduling algorithm and prove that with extra resources, it is competitive against the optimal global algorithm in minimizing the average flow time of the jobs.