| Abstract: | 經過壓縮後的視訊資料量變化極大,因而引發諸多關於如何有效利用網路資源來傳送的研究。就資料大小而言,事實上有規律可循,如果善加利用此規率性,可控制視訊的傳送時間點,以某種特定的時序,彼此交錯,我們稱之為「視訊多工(video multiplexing)」。如此一來,視訊彼此間互補資源的需求,進而提升資源(頻寬與暫存區)的使用率,同時能亦能無妥協地滿足視訊所要求的服務品質。因此在本文中,我們將此觀念運用在多節點的廣域網路傳輸,提出一套資源預留的演算法,來達成如何善用上層應用程式所允許的延遲上限,給予各網路節點適度的延遲彈性,以使他們能達到最佳的「視訊多工」效果。為了驗證所提方法的正確性與優越性,我們以Simscript模擬軟體模擬系統運作。從實驗結果顯示,當視訊流的數量大時,我們明顯的觀察到「視訊多工」所帶來的好處,一般頻寬使用率能夠達到60~70%之譜,而額外暫存區的需求亦十分有限。
It is well known that compressed videos, exhibiting significant rate variability, are hazardous to the network resource utilization. In fact, traffic also exhibits a regular pattern. If we multiplex video streams with a specific temporal relationship, they will compensate with each other on their resource requirements. As a result, the resource utilization is improved without scarifying any video quality. We term such a technique video multiplexing. In this paper, we demonstrate how to apply video multiplexing over multi-hop wide-area networks. A resource reservation mechanism is proposed to determine the transmission schedule in each intermediate hop for achieving the best multiplexing effect subject to the end-to-end delay bound tolerable to the upper-layer applications.
We simulate the system in Simscript simulation package and observe that when the number of video streams are many, the performance gain brought by video multiplexing is evident. In general, a 60~70% utilization on bandwidth can be achieved with moderate buffer space deployed in intermediate hops. |
