於個人網路提供統計式服務品質保證之研究

Abstract

由於近幾年來，穿戴式裝置與生理感測裝置等各種不同類型的行動連網感測 裝置的出現，許多更新且更多樣的應用也應運而生。 然而，隨著個人擁有的裝置 數量增加，網路的狀況會因為裝置的互相競爭造成擁擠，造成使用者得到的網路 服務品質下降。因此，如何在利用網路資源，又能讓使用者能夠達到可以滿意的 服務品質，是我們在這篇論文研究的重要議題。 在本論文中，我們討論以個人路由器作為所有行動裝置連網之核心，讓行動 連網感測裝置經由短距無線網路連結到個人路由器上，因為路由器同時擁有遠距 行動網路之介面，因此行動連網感測裝置可以經由路由器所連接的行動網路連接 上網以及連結到雲端。在此論文中，我們關注在行動連網感測裝置與路由器傳輸 的服務品質之議題。我們提出一個統計式的服務品質保證以滿足使用著的要求並 且提高網路使用效率。 在無線通訊的環境下，環境的擾動必然造成傳輸的不穩定性，因此要提供一 個精確的服務品質保證是個挑戰性的議題。在此研究中，我們套用Yaron與Sidi提 出的統計式服務品質模型，並考慮無線網路的碰撞情形加以修改。針對無線網路 環境的擾動我們使用EBF模式去描述它，傳輸的突爆流量我們則使用EBB模式去 加以描述。在這些假設前提下，我們利用排隊理論的概念得到理論值，此理論值 讓我們在無線環境的狀況下，仍能提供了統計式服務品質的保證。 本論文的理論推導結果經由實際的網路測量與ns3模擬進行驗證，已驗證在數 種個人網路環境中，我們均能提供有效率的統計是服務品質保證。 最後藉由2種 允許進入演算法，我們可以有效處理無線環境被過量環境或交通嚴重干擾之影 響，而讓此通訊系統仍可效率的被使用。

In recent years, many wearable devices and sensor devices emerge drastically, and more and more new applications show up. However, due to the increasing number of devices that one person may carry in a limited range, the network condition will be interfered because of their channel contentions, which is possible to seriously damage the QoS of the network. Consequently, how to utilize network resouce effi- ciently to provide a QoS that user satisfies is the most crucial issue that this thesis will investigate. In this thesis, we propose a network architecture with a personal router as the center of personal devices. The personal router offers Internet assess for personal devices over WiFi technology. In this way, the personal devices are able to connect the remote cloud via personal router. In our research, we focus on the QoS between personal router and personal devices. Provisioning of Quality of Service (QoS) in personal area sensor networks is challenging due to unique traffic flows of sensor applications and the fluctuation behavior of wireless capacity. In this thesis, we propose a novel approach to model stochastic properties of both wireless channels and sensor traffic processes. We employ a stochastic performance bound called the exponentially bounded fluctuation (EBF) to model the wireless channel. For sensor traffic processes, we employ the so called exponentially bounded burstiness (EBB) to characterize their stochastic arrivals. We then modify Yaron and Sidi’s method to predict the probability of satisfying the targeted delay guarantee. Our prediction is validated via measurements in a real IEEE 802.11ac network environment as well as observations via ns-3 simulations. We include the 802.11ac collision factor in the model and then implement two types of call admission controls to avoid traffic overload. In the final result, we demonstrate that the call admission control mechanisms and our model can be employed to effectively provide stochastic delay guarantees in typical personal area sensor networks.