| Wireless long distance(WiLD)networks based on IEEE 802.11 provides low-cost network connectivity for rural areas,which has great significance in applications.The bit rate of IEEE 802.11 n reaches 600 Mbps at maximum,which brings large improvement for the network performance.However it suffers poor performance in long distance links.The experiments show that the current RA(rate adaptation)schemes cannot make full use of features of 802.11 n,leading to low thoughput.RA algorithm for long-distance links is needed.IEEE 802.11 n support many MCS(Modulation and Coding Scheme),which has different performance in different links.According to extensive experiment measurement on long-distance links,we found that RSSI(Received Signal Strength Indicator)can be indicator for link status,which is also consistent with communication theory.Therefore we propose RainbowRate,a novel RA mechanism to make 802.11 n perform well in Wi LD networks.RainbowRate leverages the transition window unique in long-distance environment to cut down the set of candidate MCSs efficiently.It selects candidate MCSs according to the RSSI,which makes it react to channel variation quickly.It adjusts the bit rate and FA level by probing MCSs only in MCS candidates set and its FA level,which protects the probing frames from high loss rate since the FDR of the MCSs in the set is over 90%.Fast switch and reliable-bit-rate retransmission further ensure low loss rate and few retransmission,thus improves the link utilization.We implement and evaluate RainbowRate in our testbed.Experiment results show RainbowRate outperforms MiRA,Ath9 k and Minstrel in all the scenarios we tested.It improves UDP throughput by 100.4%,58.1% and 20.1% compared to MiRA,Ath9 k and Minstrel respectively,and improves TCP throughput by 153%,38.5% and 31% respectively. |
