On analysis of traffic flow demultiplexing effectiveness

It is known that the demultiplexing of the individual traffic flow into several independent transport subflows can increase the speed of it. This statement is true for a single flow but its truth for the massive case, when demultiplexing technics are applied to all traffic flows in the network of a single Internet Service Provider (ISP), is not obvious. The question arises, what impact the massive demultiplexing of traffic flows will have on the whole ISP network bandwidth. In this paper, this question is considered for the static case, when each flow is demultiplexed statically, i.e. every flow before its launching is demultiplexed into the same number of subflows. We developed a mathematical model that was used to construct a simulation model in order to obtain more accurate estimates of the network performance with and without flow demultiplexing. Using simulation model, network properties are defined, under which the demultiplexing of traffic flows is justified. We proved the correctness of the obtained results by emulating a network load based on a protocol stack virtualization for the same input. We considered various routing policies that can be used for massive demultiplexing. Special attention is paid to algorithms that allow you to build routes with minimal intersections, since using the nonintersecting routes with non-optimal cost can increase the network performance. The routes constructed with these algorithms were used both for the network performance analysis with demultiplexed flows and in the case of balancing non-demultiplexed flows.