Stability of Network Congestion Control with Asynchronous Updates

For network congestion control with many flows, current theoretical models predict worse performance than what the empirical study shows. In this paper, we introduce asynchronism among many flows to the traditional fluid-based model and develop an asynchronous model for network congestion control where each end-user sees slightly different snapshots of the network. Based on our model, we prove that the system is always stable when there are many flows with asynchronous updates. Both our theoretical and numerical results confirm that the asynchronism plays a dominant role in the stability of network congestion control and our asynchronous model reflects the reality far better than any other traditional fluid models with synchronous updates