無線感知雲端網路資源分配平行化與可擴性之研究

Abstract

在資訊爆炸的年代，隨著網路中資料量不斷的增加，頻譜資源逐漸供不應求。而「無線感知」技術，可以用來改善頻譜資源不足的問題。 近年來隨著無線感知技術的崛起，各式各樣的無線感知網路架構被提出包含了分散式無線感知網路、集中式無線感知網路，許多架構甚至採用了雲端來滿足大量的運算需求。 在這篇論文中，我們著重於大尺度的頻譜資源分配問題，為了達到排程的時間限制需求，我們將問題區分為兩個子問題:分群問題與頻譜資源分配問題，針對兩個不同的問題我們分別提出最佳化演算法以及時間複雜度較低的雲端可擴演算法，分群問題的目標為，在給定伺服器負載的限制下對無線感知網路基地台做分群，每一群交由一台伺服器處理。另一方面，資源分配問題則是著重在邊界無線感知網路基地台的資源分配，期望對頻譜做最佳化的利用。 在這篇論文中，我們不單單闡述問題並且設計解決問題的方式，更進一步的我們提出一些理論及定理，並且為所設計的方法分析其時間複雜度，最後，經由模擬結果以實際的數據分析並驗證其效能及可能性。

With the network in the era of information explosion, the amount of data size continuously increases and thus gradually causes the insufficiency of bandwidth resources. Cognitive Radio technique has been proposed as a promising solution to improve the spectrum utilization efficiency in wireless communications. There have been lots of researches and development in this are over the past decade. Specifically, several cognitive radio network architectures are proposed and they could be classified into distributed cognitive radio networks or centralized cognitive radio networks. Some architecture even utilizes cloud to support the massive computing demands. In this thesis, we tackle the problem of large-scale resource allocation. To meet the requirement of scheduling response time, this problem is divided into two subproblems: clustering and resource allocation. We first formulate the problem and develop an approach to have the optimal solution. Further, we propose a low-complexity scalable cloud-based resource allocation algorithm. The objective of clustering is, based on the predefined computation load constraint of each virtual machine (VM), to partition all access points (APs) into several clusters, and thus each cluster is served by a VM. On the other hand, the designed channel assignment algorithm is only performed for all boundary APs to achieve maximum channel reusability. We not only formulate the problem and design feasible mechanisms, but also propose some lemmas and theorems, and analyze the time complexity for all proposed methods. Through simulations we validate and evaluate our methods and show the effectiveness and feasibility of our approaches.