網路多媒體服務之使用者經驗品質量測與提供

Abstract

隨著寬頻網路的快速發展與高度普及，網路多媒體服務（例如網路電話和線上遊戲）已經成為許多使用者在日程生活的一部分。另外，科技的進步使得各式運算裝置的效能大幅提升，越來越多的使用者透過平板電腦和智慧型手機連結網際網路，已經不只侷限於桌上型電腦或筆記型電腦。由於運算裝置效能和網路環境的異質性，讓使用者在任何運算裝置和網路環境上都能獲得良好的服務品質，成為網路多媒體服務提供者的重要目標。然而，服務提供者必須先了解如何的服務品質才能夠滿足使用者的需求。 我們在這篇論文探討如何量測網路多媒體服務帶給使用者的經驗品質，以及探討系統層面的效能因素如何影響網路多媒體服務的品質。首先我們提出基於群眾外包概念的實驗架構，能夠有效且可靠地評估使用者對於多媒體服務品質的經驗品質與滿意程度。接著我們嘗試了解網路電話系統如何透過調整收話端的撥放延遲機制，提供使用者良好的經驗品質。另外，由於現今的網際網路無法保證資料傳輸品質，我們提出一個能夠偵測點對點檔案分享流量的機制，可用在網路管理上以提升網路服務效能。 我們的主要貢獻可分為三點：（一）運用我們提出的群眾外包式多媒體服務的經驗品質量測架構，可以有效率和經濟地從廣大受試者群體收集實驗資料，並且能夠保證實驗結果的可靠度。（二）我們探討三個熱門的網路電話系統的撥放延遲機制，並指出這些系統可藉由改善它們的機制而提供使用者更好的服務品質。（三）我們提出的點對點檔案分享流量偵測機制能夠達到很高的偵測率，能夠對於網路服務品質提供帶來很大的助益。

Due to the ubiquity of broadband Internet access, more and more people enjoy networked multimedia services, such as online gaming and VoIP, in their daily life. With the prevalence of powerful mobile devices, an increasing number of people use tablet computers or smart phones—rather than traditional desktop and laptop computers—to access the Internet. Because of the heterogeneity of hardware capability and network environment, providing users a satisfying experience no matter what platform they are using is the ultimate goal of service providers. However, service providers need to know what level of the quality of multimedia services is required to satisfy users. In this dissertation, we endeavor to explore how to assess QoE (Quality of Experience) of networked multimedia services and how QoS (Quality of Service) factors affect QoE of networked multimedia services. First, we propose a crowdsourceable framework to quantify the QoE of multimedia content efficiently and reliably. We then investigate how a VoIP application controls its playout buffer, i.e., one of the most important QoS factors of VoIP, to provide users with satisfactory QoE. Since the Internet is not QoS-guaranteed, we then propose a P2P traffic identification scheme that is useful for improving network QoS. Our main contributions are three-fold: 1) Our proposed crowdsourceable multimedia QoE assessment framework help researchers efficiently and economically obtain QoE experiment data from a diverse labor pool without compromising the reliability of the assessment result. 2) We investigate the playout buffer dimensioning algorithms of Skype, Google Talk, and MSN Messenger and demonstrate that these VoIP systems could provide users with better QoE by improving their algorithms. 3) The proposed P2P traffic identification scheme achieves high detection accuracy and thus is very helpful for network QoS-provisioning.