End-to-end Performance Optimization of Tandem Queuing for High-Speed Train Networks

Train passenger communication services have gained more and more attention due to the enormous growth of multi-media services, as well as the development of high-speed railway. In order to deal with the high outage possibilities caused by penetration loss and signaling storm of group handover when communicating with the wayside cellular network directly. The High-Speed Train (HST) WiFi network, which provides passengers' information access within the carriages, emerges at the moment. However, there are still some issues which are not well addressed. With more users accessing to the WiFi network, the onboard relay connecting both of wayside cellular network and carriage WiFi network becomes a bottleneck, which significantly affects the network performance. Furthermore, by abstracting each carriage as a sub-network, the HST WiFi network can be seen as a tandem multi-hop network which will raise other challenges such as the end-to-end error accumulation and the bandwidth progressive loss carriage by carriage. In this paper, we obtain the arrival process of each sub-network by using a decomposition method and a parameter fitting algorithm. Then, we model the sub-network as a discrete-time finite capacity queuing system with packets splitting by a Markov Modulated Bernoulli Process (MMBP)/G/1/K queue. In addition, we use Hybrid Automatic Repeat Request (HARQ) in each sub-network to relieve the end-to-end error accumulation. Through the analysis of the stationary queue length distribution, average delay and packet loss probability, we obtain the optimal buffer size to ease the traffic congestion and give some suggestions on resource allocation of HST WiFi network.