TCP performance over cdma2000 RLP

Abstract

This paper provides performance results for TCP over cdma2000 radio link protocol (RLP). The simulations are performed for a single user running a long FTP application with "infinite" data backlog. With the assumption of a long FTP application, maximum TCP throughput is the relevant optimization metric. The goal is to analyze the impact of TCP window size, RTT, wireless channel bandwidth, FER model and RLP transmission delays on end-to-end transport. The IS-95 and its evolutions (i.e. IS-95B and cdma2000) use a NAK-based RLP. The number of rounds of NAKs and the number of retransmissions in each round is varied. We analyze the impact of varying different RLP parameters and retransmission schemes on TCP throughput. Our findings demonstrate that the (1,1,1,1,1) NAK scheme outperforms the (2,3) NAK scheme when the errors are correlated. This is also true for non-correlated error case and larger TCP window size. However, with non-correlated errors and comparatively smaller TCP window size, the (2,3) NAK scheme performs better. It was also observed that the TCP throughput is not severely penalized up to link error rates of 10% (approximately 15% loss in TCP throughput at 10% FER).