IEEE 802.11e HCCA中改善變動位元速率資料傳輸效能上鏈排程演算法之研究

Abstract

在IEEE 802.11e HCF (hybrid coordination function)提供以集中式輪詢(polling)機制來存取通道，稱為HCCA (HCF controlled channel access)。存取點(access point, AP)透過排程機制(scheduling)，將可提供即時多媒體影音資料的服務品質(QoS)保證。為了改善上鏈傳輸變動位元速率(VBR)資料流的效能，過去有不少學者提出研究，例如ARROW排程演算法，此方法藉由回授方式將資料累積量上傳給QAP來給予精確的頻寬分配，但是因回授與分配頻寬為不同時間之動作，產生的傳送延遲將造成資料封包超過延遲時間限制而遺失。因此在本論文中，將提出一個四路輪詢排程演算法。在輪詢上鏈QSTA時，此排程器將會直接詢問緩衝器情況，並且立即給予精確的頻寬分配。同時論文中也會提出數學模型來對四路輪詢演算法和ARROW進行數值效能分析，也將藉由ns-2的實驗模擬將會看到本論文作法相較於ARROW除了在平均封包遺失率上可有效地降低，同時也能得到更大的容量並滿足多媒體影音服務品質。 而在IEEE 802.11e HCCA中，同時也需考慮系統的允入控制，藉由允入控制功能可計算出系統容量，並且給予進入連線所要求的服務品質。對於VBR類型的資料流，學者提出RVAC (Rate-Variance-envelop-based Admission Control)方法，利用Dual Token Bucket機制將進入系統的資料流塑形外，同時將統計多工增益的觀念引入HCCA。但在此方法中，並未考慮不同資料流的延遲限制差別與實際排程之情況，喪失能取得較大多工增益之效能。故我們將提出一個在Rate-Variance-envelop基礎下針對VBR之高效率動態允入控制的設計想法，將考慮各別資料流延遲特性，動態調整允入方式讓系統頻寬消耗較少、得到較大的多工。最後，由分析結果將會看到本論文作法相較於RVAC能讓整體效能改進，有效提升系統容量。 In IEEE 802.11e, a centralized polling based channel access mechanism is provided for the quality-of-service (QoS) provision of real-time applications. In order to improve bandwidth efficiency of polling for uplink variable bit rate (VBR) sources, it has previously been proposed in, e.g., ARROW, that the amount of backlogged traffic is fed back for exact bandwidth allocation. However, due to that the feedback and the bandwidth allocation are performed separately in two different polls, transmission latency incurred by packets could cause them to violate delay constraint. In this paper, a four-way-polling scheduler is first proposed by directly inquiring buffer occupancy information of an uplink station during a poll. Theoretical analysis is then performed to evaluate the performance of both the proposed scheduler and ARROW. From both numerical and ns-2 simulation results, it is shown that packet loss rates can effectively be reduced and more capacity can thereby be obtained under the proposed scheduler, as compared with ARROW. Besides, there is also a need to consider the admission control of VBR traffic over HCCA. RVAC has previously been proposed to improve network utilization. In RVAC, Dual Token Bucket shaper is applied and the existing statistical multiplexing framework is proposed to design admission control scheme. However, RVAC fails to consider the characteristics of distinct VBR traffic flows, so it cannot satisfactorily exploit statistical multiplexing gain. In this paper, we design an admission control scheme that effectively acquire additional multiplexing gain based on Rate-Variance- envelop to deal with different delay requirements of VBR traffic flows. From numerical results, it is shown that more capacity can be obtained under the proposed admission control scheme.