在網際網路隨選視訊服務之快取伺服器中提供一個有效置換資料方法的研究

Abstract

由於電腦與多媒體技術的進步，隨選視訊服務以被廣泛的利用。但受限於現階段的網路技術,往往使得網路頻寬成為整個視訊伺服器設計的瓶頸。為了解決這個問題，在這篇論文中，我們提出了一個視訊快取伺服器的架構。若使用者觀看相同的影片時，我們將資料存在近端的視訊快取伺服器中，可以大量的減少隨選視訊伺服器到視訊快取伺服器這一段的頻寬。而視訊快取伺服器擁有高命中率才能真正省下最多的頻寬，因此，我們設計了一個有效置換資料的方法。此一方法之特點是利用knapsack演算法，讓視訊快取伺服器能存放最熱門的影片以減少最多的頻寬使用量。除此之外，我們還提出了一套可以預測今天使用者分佈的預測方法，藉由這個方法，我們可以使視訊快取伺服器達到最高的效能。從模擬結果我們可初步的分析出，我們的方法可以省下大量的頻寬，進而減低網路高負載的情況。詳細的設計原理、運作流程以及效能評估將會在論文以後的章節中一一介紹。

Due to the advances in computer and multimedia techniques, the application of video-on-demand is popular. However, restricted by the network technology, the bandwidth becomes a bottleneck in the VOD environment. In order to resolve the problem, we will propose a video proxy architecture in this thesis. If many clients watch the same program, caching the data into the video proxy can save huge bandwidth from the remote home sever to the local proxy. A good video proxy has higher hit rate. Thus, we design an efficient data replacement scheme to increase hit rate. We use knapsack algorithm to cache the most popular videos in our proxy so as to reduce the most bandwidth usage. Beside, we also provided a prediction method to predict the distribution of access patters. By the method, we can raise the total performance of the video proxy. According to our simulation, our method can save a lot of bandwidth so as to lighten the network load. The detail information about design principle and performance evaluations will be described in the literature.