點對點的多媒體近似直播串流儲存策略

Abstract

點對點架構(peer to peer)是在application層建立協定，讓網路上的peer能遵從一樣協定來進行溝通與傳輸資料。每個peer都可像主從式架構的客戶端去尋找資源，也可像伺服器提供資料，回應他人需求。此篇論文模擬在網路的應用層上，建立p2p的協定，利用像[3]的mesh-pull架構，做直播的多媒體串流分享，並提出peer之間合作式的cache儲存協定，讓使用者可以選擇收過去播過的多媒體串流，不只侷限在直播的串流。我們稱此為「近似直播串流(Near-live streaming)」應用，其結合了直播(live streaming)與隨選播放(VoD)功能於系統中。 一般直播多媒體系統多只儲存與直播節目最近期的資料，其cache資料更新採用FIFO的方法(或稱continuous Overwriting)，此種方法雖然簡單卻在所提的近似直播的應用中，普遍有較差的效能。所以我們提出No overwriting的caching，與前述方法做互補。為了適合更多數的video popularity的情況，提出Adaptive overwriting的方法。實驗結果顯示所提的caching方法，特別是Adaptive overwriting，在大多數的情況下昆能有較好的效能。

The peer to peer communication is a protocol for user sharing computer resource and service over the application layer in the Internet. Every peer acts client and server, there is no central server. This paper extends the P2P protocol at live media streaming in [3] and proposes several cooperative caching strategies to provide “Near-live streaming service” which combines live streaming and VoD service in P2P systems. In these caching strategies, the simple and general method is continuous overwriting replacement, but it shows low hit rate in most of the cases for providing VoD services in the near-live streaming application. Therefore, we purpose a method called “No overwriting” caching, it mends drawbacks of the overwriting continuous method. But the “no overwriting” method still can’t satisfy the variable demand for users on large scale p2p system. Since this, we purpose an “adaptive overwriting” method that can suitable most of video popularity distribution. The simulation results demonstrate the superiority of the proposed methods.