A New TCP Congestion Control with Weighted Fair Allocation and Scalable Stability

Recent studies on end-to-end congestion control in IP networks have demonstrated the difficulties in compromising between stability mid fairness objectives. In this paper, we introduce a new congestion avoidance model, which can meet the key criteria of both objectives with enhanced source behavior. Our new scheme provides local system stability which is scalable to arbitrary delays, capacities and network topologies. The tradeoff between system stability and response speed is controlled by single parameter tuning at source. At equilibrium, the obtained resource allocation is weighted proportional fair or weighted TCP friendly fair with appropriate choice of parameter at source. Not like the arrangement in current TCP scheme, in which large round-trip time (RTT) are treated with bias, our resulting weighted allocation scheme favors long delay flows