Congestion control and switch buffer allocation in high-speed networks

Congestion control and switch buffer allocation techniques are proposed that allow for packet admission and transfer based on the switch and network occupancy. The packets are transferred if the required bandwidth is smaller than the bandwidth currently available. Otherwise, the packets are stored in the buffer. If the buffer is full, the incoming packets are dropped and the transmission is delayed either at the sender or at the nearest node. It is proven analytically that the packets will not be dropped if the buffer size is greater than or equal to twice the size of the queue length and the traffic will not increase during propagation delay of shut-off signal between the switch and either the source or the nearest node, depending on where the transmission is delayed. The performance of different networks is evaluated to show the applications of the proposed techniques.