TCP window based dynamic assembly period in optical burst switching network

In OBS network, assembly period and burst drop rate impact the performance of the upper layer. This paper evaluates the effect of parameters of OBS network on the throughput of TCP. Based on a general network model, we made theoretical analysis and simulation. In the analytical model, a parameter called window data dropping probability p(w) is introduced to study the correlation between the TCP window size and parameters of OBS network. Both theoretical and numerical results show the same trends, which verify the correctness of the analytical model. From the results, an optimal assembly period is found to exist under certain burst drop rate, with which we can obtain the best throughput of TCP. Dynamic Assembly Period (DAP) algorithm we proposed can track the variation of the current TCP window size and update the assembly period dynamically for the next assembly operation. Simulation results show that DAP algorithm can improve the performance of TCP over OBS network.