Achieving distributed buffering in multi-path routing using fair allocation

Traditional network routing algorithms send data from the source to destination along a single path. The selected path may experience congestion due to other traffic, thereby reducing throughput. Multi-path routing divides the data along multiple paths, with the aim of increasing overall throughput. Due to the different latencies of using multiple paths, packets are received out-of-order at the destination, and thus require reordering. This problem can be overcome by placing a buffer at the destination node in order to synchronize the incoming packets across the multiple paths. However, this solution is not scalable, since using a single large buffer is impractical, and can be expensive. An alternative is to distribute the buffering requirement along bounded buffers located at the intermediate nodes on the path, thereby reducing the need for a large buffer at the receiver. This work proposes several fair allocation algorithms aimed at improving the existing methods already in place to distribute the buffer requirement across the network in an efficient manner.