無線網路下基於適應性加權之公平排程演算法

Abstract

無線網路與有線網路的公平排程有很大的不同，因為在無線網路中有位置相關以及突發性的頻道錯誤發生。無線傳輸媒介遇到相當程度的鏈結錯誤時會導致鏈結容量損失。此外，負責對上傳與下載頻道做排程的基地台對行動主機的頻道狀況只有有限的瞭解。在這種情況下，針對無線網路提供服務品質是一大挑戰。在本論文中，我們提出一個使用適應性加權的公平排程演算法(AW-FQ)，它可以對遇到頻道錯誤的資料流提供較好的服務品質，而且不需要預測無線環境下封包傳輸之頻道品質。本方法會依據所遇到的封包錯誤率動態調整資料流的權重。當發生頻道錯誤時，每個資料流都能保證收到其原有服務的最小可容忍品質。因為無線環境的傳輸品質並不穩定，AW-FQ特別適用於需要不同最小服務品質要求的應用。在AW-FQ中，每個資料流都被分配一個不總是佔住部分固定頻寬的虛擬資料流。此虛擬資料流可以對其他無錯誤發生的資料流提供額外服務，也能對它所對應的資料流進行補償。當一個資料流提供補償時，其服務量的降低可以控制在事先預定好的可容忍範圍內。模擬結果顯示，我們的方法可以保證傳輸量與公平性。與SBFA比較，AW-FQ有較佳的佇列延遲範圍，同時也不需要指定會佔用部分固定頻寬的資料流來提供補償。

Fair scheduling in wireless networks is significantly different from that in wireline networks because of location-dependent and bursty channel errors. The wireless media may experience a substantial rate of link errors that cause the loss of link capacity. In addition, the base station which performs scheduling for both uplinks and downlinks has only limited knowledge of the mobile host's channel conditions. This results in a great challenge to provide quality of service (QoS) in wireless networks. In this thesis, we present an adaptive weighted fair queueing algorithm (AW-FQ) which can provide better QoS for error-prone flows and does not need to predict the channel quality of packet transmission in the wireless environment. The weight of a flow is dynamically adjusted according to the experienced packet error rate. Every flow is guaranteed to receive the minimal tolerable quality of its original service when experiencing channel errors. Since the wireless environment is unreliable, AW-FQ is suitable for applications that need different minimal QoS requirements. In AW-FQ, each flow is individually assigned a virtual flow which does not always occupy a fixed fraction of the bandwidth. The virtual flow can provide additional service for other error-free flows and compensate the flow which it is assigned to. The amount of service degradation of a flow can be administratively controlled within a predefined tolerable ratio when the flow provides compensation. Simulation results show that our approach can guarantee throughput and fairness. Compared to the server based fairness approach (SBFA), AW-FQ has better queueing delay bound and does not have to assign a fixed flow which always occupies a fixed fraction of the bandwidth for providing compensation.