串流平台多使用者的傳輸時間分析

Abstract

串流伺服器因為必須同時服務多個用戶的連線需求以及要滿足即時傳送 封包給不同用戶，因此如何設計一個順序來控制每張frame的傳送時間是本論文重點。

伺服器會對每一張frame計算其PTS(display order)和傳送時間DTS(Sending order )，並且依照DTS(Sending order)來傳送該frame。當某個使用者的frame所切割出來的payload的DTS(Sending order)小於系統時間(Wall clock)時，意即該payload可以開始經過封裝傳送的處理，而在此需特別強調的是同一個frame所切割出來的payload都會具有同樣的DTS(Sending order)。

本論本研究的核心所在便是去研究串流伺服器如何使用單一行程來達到多工，其時間排程的機制、以及設計理念。

Streaming servers need to consider the simultaneous transmission requirement for all users, thus how to design a mechanism to control every packet’s transmission time will be a critical issue in the server design.

For real time transmission, server would calculate a sending order for each frame. This order is denoted as DTS. The packets associated with a frame are sent according to DTS of a frame. During transmission the DTS of a payload of one user is compared with the Wall clock and packets are sent sequentially.

The thesis is to discuss the method that can be used to maintain a smooth transmission order for all frames in a multi-user environment such that the real time transmission can be achieved and the network is not over-flown.