IEEE 802.11速率控制演算法在MADWIFI平台上的實作

Abstract

近10年間，IEEE 802.11無線區域網路儼然成為世界上最熱門的無線通訊技術。IEEE 802.11標準在實體層（physical layer，PHY）中提供了多種傳輸速率，例如：802.11b可支援4種速率，而802.11g則提供多達12種速率。不過，它並無指定速率調適的機制，而此開放式議題稱為自動速率控制（auto rate control）或速率調適（rate adaptation）。在隨機無線通道環境下，有效地決定一種合適的傳輸速率顯得特別關鍵。本篇論文中，在開放程式碼的Linux核心（kernel）驅動程式─MADWiFi平台，我們實作了一種新穎速率控制演算法。首先，我們追蹤MADWiFi的程式碼，以得到一些有用的函式及變數。接著，我們在一段估測時間中，收集訊框（frame）傳送成功與否的資料，來計算每種速率的訊框成功率（Frame Success Rate，FSR）。最後，所估計的FSR值對應到以一維馬可夫鏈模型（one-dimensional Markov chain model）為基礎的傳輸時間分析。由於設計新穎速率控制演算法是為了達到最大的傳輸量（throughput），所以我們選擇一個具最少傳輸時間的速率來做為下一次訊框傳遞的速率。透過真實世界中不同通道和拓樸（topology）環境的測量，實驗結果證明：在所有測試情境下，新穎速率控制演算法勝過在MADWiFi現有的4種演算法（即AMRR、ONOE、SampleRate及Minstrel）。值得一提的是，在戶外高速移動情況下（10m/s），我們的實作在傳輸效能上至少有51％的顯著改善。

In the last decade, the well-known IEEE 802.11 wireless local area networks have become the most popular wireless communication technology in the world. IEEE 802.11 standard provides multiple data rates at physical layer (e.g., 802.11b supports four rates, while 802.11g offers twelve rates). Nevertheless, it does not specify a rate control mechanism, and this open issue is called 'auto rate control' or 'rate adaptation'. How to efficiently determine an appropriate transmission rate is a crucial issue, especially in stochastic wireless channel condition. In this thesis, we have implemented a novel rate control algorithm on the MADWiFi platform which is open-source Linux kernel driver. First, we trace the source code in MADWiFi, and some useful functions and variables are obtained. Next, we collect frame-deliveried information in an estimation window to calculate the Frame Success Rate (FSR) for each rate. Finally, the estimated FSR values are mapped to the analytical transmission time based on one-dimensional Markov chain model. Since the novel rate adaptation algorithm is designed to achieve the maximum throughput performance, a specified rate with the smallest transmission time is selected for next frame delivery. Through extensive real-world measurements under different channel and topology environments, experimental results demonstrate the novel rate control algorithm outperforms four existing algorithms on MADWiFi (i.e., AMRR, ONOE, SampleRate and Minstrel) in all tested scenarios. The worthiness of our implementation has significantly enhanced the throughput performance, at least 51%, in outdoors with high-speed mobility (i.e.,10 m/s).