Analytical Model of Adaptive Cross-Layer-Based TCP Congestion Control for Wireless Mobile Cloud Access

Abstract

In cloud computing, the all-IP Internet acts as the highspeed wireless mobile network for the smart devices (e.g., smart phones, smart PADs, etc.) to access the plentiful cloud services. The data-packets of cloud applications are transmitted via the reliable end-to-end TCP or the unreliable UDP protocol. For reliable cloud transmissions, although TCP can operate well in a wired network, it suffers from wireless high-interference and high packet loss rate in heterogeneous wireless networks. The main reason is that the wired network congestion causes consecutive packet loss but the wireless link error leads to inconsecutive packet loss. This paper first proposes a cross layer mechanism to differentiate whether a loss event is caused by the wired network congestion or by the wireless link error. Then, a Markov chain-based analytical model is proposed to analyze TCP four states probabilities of the adaptive TCP approach. Finally, we analyze the throughput and the TCP growth function of the proposed approach. Numerical results demonstrate that the analytical results are very close to simulation results under different loss rates, which validates the proposed analytical model achieving excellent features. The proposed approach outperforms other TCP approaches in goodput while yielding competitive fairness.