自組織雲端網路中雲端的可擴充性與資源分配之研究

Abstract

近年來，隨者網路使用者不斷的增加以及使用者的上網習慣也在逐漸改變，越來越多使用者利用網路來觀看影音服務，導致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。 但是如果想要服務數以萬計的使用者，那麼勢必需要一控管中心，在我們的架構中我們以雲端的方式去集中管理我們的無線感知網路基地台以及使用者。在這篇論文中我們著重於兩個部份，第一是對於雲端系統架構的討論，要如何才能使我們的系統更好，像是如何針對雲端中各個元件去做擴充，以及對於雲端中的資料傳遞。第二，考慮一大型環境，我們為了支援即時的通訊服務，我們提出分群演算法將一個大型環境切割成數個子群，而每個子群都要滿足所限制的工作負載量，並且將每個子群分給一台虛擬機器負責分配頻譜資源，如此一來就可以帄行的分配各個子群的頻譜資源。而當我們分出子群之後，對於子群與子群間邊界上的無線感知網路基地台會有干擾情況發生，而不同子群為不同虛擬機器所管理無法立即得知彼此對於頻譜的分配情況，所以我們提出邊界頻譜資源分配演算法對邊界上的無線網路基地台頻譜先做一個初步的頻譜資源分配。 對於分群與邊界頻譜資源分配問題，我們分別提出的最佳化與啟發式演算法並做時間複雜度與模擬的比較，並且會針對一大環境做模擬來驗證我們的雲端系統的可行性。

In recent years, Internet users continue to increase as well as the user's surfing habits are gradually changing,, more and more users use the Internet to watch video services, cause the shortage of spectrum resources. The cognitive radio technology can be used to improve the spectral problem of insufficient resources. But if we want to serve tens of thousands of users, it is need a control center, in our architecture we use cloud to manage our cognitive radio network base stations and users. In this paper we focus on two parts , the first is for the discussion of cloud systems architecture , how to make our system better , like how to scale-out each component of the cloud , as well as in the cloud data transfer . Second, consider a large-scale environment, we have to support real-time communication services, we propose a clustering algorithm will be cut several cluster in a large-scale environments, and each cluster must meet the work load is limited and will be allocated to each cluster a virtual machine is responsible for the allocation of spectrum resources, as a result, we can parallel the distribution of the spectrum of each cluster. But when we separate the cluster, the boundary base station between cluster and cluster will cause interference, and different cluster management by different virtual machines can not immediately know each other for spectrum allocation situation, so we propose a algorithm for wireless networks for boundary base stations do a preliminary allocation of spectrum resources. For clustering and boundary base stations spectrum resource allocation problem, we were presented optimization and heuristic algorithms and do the time complexity compared with the simulation, and will make for a large-scale emulations to verify the feasibility of our cloud systems.