多媒體資料流傳輸系統的強韌性及流量控制機制設計

Abstract

錯誤控制和流量控制是多媒體資料在網路上傳輸的關鍵技術。 為了克服頻寬的變化和傳輸的不可靠性，先進的多媒體傳輸系統通常會以封包的丟失為依據，以此來提供錯誤和流量控制機制。雖然封包的丟失在有線網路的Internet上是網路緩衝滿溢的徵兆，但對於無線網路而言並非正確。一旦觀察到有封包丟失的現象，流量控制機制可能會馬上嘗試著減少傳輸的速率。然而，在無線網路的傳輸上，連續的封包丟失可能來自於網路阻塞或是較高的位元錯誤率。這樣的結果，導致流量控制可能使得大部分的時間所使用的頻寬較低於通道的真實頻寬。 若流量控制的做法採取積極的方法，重新傳送的錯誤控制機制可能會真正因網路阻塞而被妨礙。在這篇論文之中針對資料流切換系統提出了一個創新的方法，這個方法結合了流量控制和以重新傳送為基礎的錯誤控制機制。資料流切換是現今使用在多媒體工業中，最普遍用於動態調整傳輸速率的技術。此外為了實用性的考亮 它還有著沒有編解碼器上的束縛的優點。不過，目前的資料流切換系統通常會遭遇之前所提到的兩個問題。這裡所提出的技術可以同時達到降低重新傳送資料的延遲和較高的通道頻寬利用率。根據實驗的結果顯示，這個新的技術對於多媒體傳輸服務是有效的。

Error control and flow control are key technologies for multimedia streaming over the Internet. In order to cope with bandwidth variation and unreliability of the transmission channels, advanced multimedia streaming systems usually provide error and flow control mechanisms based on packet loss patterns. While packet loss is a good indication of network buffer overflow over wired Internet, it is not true for wireless IP networks. A flow control mechanism may try to reduce the transmission rate as soon as packet loss is observed. However, over wireless channels, consecutive packet losses can come from network congestion as well as from high bit-error rates. As a result, the flow control algorithm may under use the channel bandwidth most of the time. If the flow control algorithm takes an aggressive approach, a retransmission-based error-control mechanism can be hindered greatly due to true network congestion. In this thesis, an innovative joint flow control and retransmission-based error control mechanism is proposed for a bitstream switching system. Bitstream switching is the most popular dynamic data rate adjustment techniques used in the multimedia industry. In addition to being practical, it also has the advantage of not constrained by the type of codecs (both audio and video) you use. Nevertheless, current bitstream switching systems usually suffer the two aforementioned problems. The proposed technique can achieve both the goal of reducing retransmission delay and making full utilization of the channel bandwidth. Based on the experiments, this new technology is very promising for multimedia streaming services.