Cooperative congestion control in multi-level InterWeave routers

Next generation data centers and cloud computing architectures require high data rates of the order of 100 Gbps from the Internet. However, most routers are typically implemented as clusters of computer and line cards, where it is difficult to allocate bandwidth and to guarantee flow fairness. In this paper, we present a cooperative congestion control scheme in the multi-level InterWeave routers with non-blocking switch fabric. The proposed scheme can provide differentiated services according to the allocated bandwidth, and can also automatically increase the bandwidth allocation of existing traffics classes. In particular, a novel AQM (Active Queue Management) algorithm is adopted in the forwarding board to guarantee the UDP flow fairness. Our simulations under two different network architectures demonstrate that the novel cooperation congestion control can achieve better fairness and data flow protection.