推拉式點對點即時影音串流系統之服務品質改善

Abstract

近年來興起的推拉式(Push-Pull)點對點影音串流系統，結合了以拉為基礎(Pull-Based)和以推為基礎(Push-Based)兩種架構的優點，被認為是能夠更有效率地將影片內容散播出去的做法。 我們發現到影片品質保證在這樣的系統之下，是一個重要的議題，而目前運作中的商業系統中，都沒有針對使用者的影片觀賞品質提供任何的保護機制，因此當網路不穩定或者頻寬不足時，影片的觀賞品質將受到嚴重的影響。然而我們發現在影音串流(Video Streaming)系統下常見的做法，重要性優先排程演算法(Importance-First Scheduling)，並不能被直接地套用到點對點(P2P)的系統下，因為當系統中使用者數量規模變大時，可能會造成嚴重的內容瓶頸現象(Content Bottleneck)。 因此在這篇論文中，我們提出了一種策略切換型推拉式演算法(Strategy-Switch Push-Pull Scheme)，讓每個節點會根據目前網路情況以及緩衝區接收資料的情形，來動態的調整目前的排程演算法，避免在低優先權資料區塊(Low Priority Chunks)上發生嚴重的內容瓶頸現象。最後，我們採用模擬的方式來驗證我們所提出的方法，實驗結果顯示，我們的方法不僅可以維持原本推拉式系統的良好效能，且同時還提供了良好的影片品質保護機制。

During recent years, the emerging push-pull P2P streaming architecture has appeared to be the most promising way to efficiently distribute video content over the Internet which offers a good trade-off between pull-based and push-based architecture. We observed that video quality guarantee is an important issue for the users in these systems, but the current push-pull p2p streaming system contains no video quality guarantee mechanism to prevent users from video quality degradation when encountering network bandwidth or resource inadequacy. However, importance-first (IF) scheduling cannot be applied to P2P streaming directly, because it may suffer from serious content bottleneck especially in large scale. In this work, we proposed an adaptive Strategy-Switch Push-Pull Scheme that peers will adaptively adjust the scheduling strategy between random scheduling and importance-first scheduling scheme according to their buffer status to prevent from serious content bottleneck for low priority chunk. We use the simulation to show that our scheme can not only keep the characteristic of high performance of the push-pull architecture but also achieves the better video quality guarantee.