隨選視訊保證服務品質的批次系統設計

Abstract

隨選視訊系統提供使用者選擇影片的服務，並提供如錄放影機的互動功能。如何有效的利用系統資源提供更多客戶享受視訊服務，是架構隨選視訊系統的重要課題之一。批次服務便是屬於這樣的一種技術，它使用資源共享的觀念，利用一份系統的資源來同時服務多個使用者，達到資源的有效應用。但批次服務產生的等待時間將導至使用者取消其視訊要求，而影響到整個系統的服務品質。一個有效的批次排程方法是批次系統設計的核心技術。在這篇論文中，我們提出一個新的批次排程方法來提高服務品質。我們將重點集中在使用者離去的行為模式，先服務可能離去機率較高的視訊要求，以減少批次系統所造成的負面影響。我們也研討及設計使用者互動下有效分享資源的方法並降低緩衝區的需求。根據模擬實驗的分析證明我們所提的方法，確實能有效的利用系統資源，提昇系統服務的效能。

A Video-on-Demand (VOD) system provides the ultimate flexibility in video services by allowing users to select video programs, at any time, and to perform VCR-like user interactions. To design a VOD server, one approach is to have a dedicated video stream for each customer. This is expensive, especially when multiple identical video streams are sent to multiple customers accessing the same video program. Batching of video requests is often used to reduce I/O demand and to improve throughput. The basic idea is to delay the user requests for a certain amount of time so that more requests for the same video arriving during the current batching interval may be serviced using the same stream. As the delay grows larger, viewers might leave the system without being served. In this thesis, we consider the user reneging behavior models and propose the Maximum Expected Loss (MEL) batching policy. We then improve the MEL policy by combining the Maximum Factored Queue Length (MFQ) scheme to become a better policy. It is referred to as the MEL*MFQ policy and performs as an excellent choice batching scheme with respect to all of the performance measures. To support user interactions under batching, we design a Split and Merge protocol with adaptive piggybacking (SAM_AP) to reduce the cost of additional buffer. The small interaction delay constraint of user interactions is a critical issue. In order to reduce user interaction delay, in our SAM_AP protocol, we maintain a small pool of reserved streams for the purpose of immediately providing user interaction requests. A simulation is developed in this thesis and simulation results show that our proposed batching system does utilize system resources effectively and improvement the overall system performance.