在無線區域網路中以群集為考量的彈性頻寬分配增進系統吞吐量

Abstract

隨著無線網路的普及性日益劇增，不論是各式各樣的可上網設備，抑或是隨處皆可使用的無線網路都讓上網變成是一件很方便的事。而在使用無線網路時，「連線品質」是讓人對無線網路產生質疑的部分。連線時發生延遲的現象，不只惱人，更容易讓人排斥對無線網路的使用。 從無線網路開始使用以來，有關連線品質的研究不勝枚舉，而本篇所要探討的是在近年開始被人討論且用來解決連線品質的相關問題：彈性的調動頻道頻寬。較寬的頻道頻寬可以傳輸的資料量較大，傳輸時干擾的範圍又小；而較窄的頻道頻寬則是可讓接收者收到的訊號較佳、傳輸時所佔用的頻段範圍較小。 了解了頻道寬度大小的特質後，下一個問題便是要如何決定存取點所該使用的頻道寬度和其中央頻率。根據過去的研究指出，互相偵測得到彼此的存取點，若使用不重疊的較小頻道傳遞時，其最終的吞吐量會比使用有重疊的較大頻段的吞吐量還多。故在分配存取點的頻道寬度時，如何讓存取點間的干擾關係降低便是我們主要的精神。一旦干擾減少了，其吞吐量也隨之增加了。 另外，有別於過去的研究只著重在頻道寬度和中央頻率，本篇加入了考慮「不正常的速率」的問題和重新探討用戶端和存取點間的連線關係。在考慮到不正常速率時，我們希望在傳輸資料時，用戶端佔據頻道的時間能盡可能的平均，而不是讓速率較慢的用戶端拖累整體系統的吞吐量。至於重新探討用戶端和存取點間的連線關係，則是因為當存取點使用不同的頻帶寬度傳輸時，其接收端的訊號強度將隨之影響，而傳統讓用戶端連結到其收到訊號強度最強的存取點的連接原則將無法再決定存取點使用的頻道寬度時先行決定用戶端所連結的存取對象。 最後，本篇在綜合上述的所有觀點後，設計了兩個演算法，其一的目標為在維持用戶端時間的公平性下最大化整體系統的吞吐量，另一個則是極大化用戶端佔據頻道的時間公平性。並將提出的演算法模擬出來後和修正過後的過去的研究相比較，不論是在系統的吞吐量和用戶端佔據頻道的時間公平性上，我們都可以明顯地優於過去的研究，顯示綜合考量所有因素而設計的演算法，對整體系統的效能有大幅度的提升。

With the increasing generality of the wireless network, various equipment which can surf the Internet and lots of place where we can use the wireless network both make surfing the Web become an easy thing. “Link Quality” is the most doubtful part for people when using the wireless network. If some delay phenomenon happened when using Internet, it not only lets people annoyed, but also eliminates their desire to use wireless network. There are a lot of researches about Link Quality when people started to use wireless network. This paper wants to discuss about “Flexible Channelization” which is used to solve the Link Quality problem in recent years. Using wider channel width can transmit more data and has smaller interference range; however, using narrower channel width can let receiver get stronger signal and occupies fewer space in spectrum. After knowing the property of various types of channel width, we want to know that what channel width and central frequency are used for each AP when transmitting data. According to existing researches, if APs which have interference relation with others use non-overlapping but smaller channels, the final throughput will be more than using overlapping but larger channels. In addition , differs with past researches which only focus on central frequency and channel width, we added the “Performance Anomaly” issue and re-discussed the association relation between APs and clients. When considering the “Performance Anomaly” issue, we hope the time of each client using the channel can be distributed for clients more fairly, and not letting the clients with slower rate degrade the total system throughput. In re-discussing the association between APs and clients part, because receiver will get different signal strength when APs using different channel width to transmit data, the traditional way which letting clients associated to the AP which provide strongest signal strength are not usable in flexible channel environment. Finally, considering all issues mentioned above, we design two algorithms: one for keeping time fairness for clients to maximize the system’s total throughput, the other for maximize the time fairness which clients using to receive data. We compare the simulation result of our works and past researches. No matter the whole system’s throughput and time fairness, we outperforms others, and it indicates that the algorithms which designed by considering all aspects mentioned above will increase the functionality of the whole system.