Macrocell與Femtocell共存系統數據流量建模與阻隔機率計算

Abstract

毫微微細胞基地台(Femtocell)與巨細胞基地台(macrocell)共存系統是將來行動通訊的趨勢之一，本研究探討在無線端之連結流量需求、資源限制與數據服務效能間的關係，利用連結允入控制機制設計以作為頻寬配置決策的基礎。 具體的研究情境為一個有頻譜執照的電信業者，擁有macrocell及femtocell共存系統，用戶接取兩子系統進行數據服務，macrocell訊務流量匯集至後端固網Iu-ps介面處，femtocell訊務流量則匯集至Iuh介面以連結網際網路。Connection Admission Control(CAC)分別由macrocell子系統的Radio Network Controller (RNC)與femtocell子系統的Femtocell GateWay (F-GW)執行，中心管理系統Admission Control for Coexistence Coordinator (ACCC)統計並管制macrocell和femtocell用戶的訊務流量。無線資源會限制住來自macrocell用戶之流量，新進之用戶連結需求會被RNC阻隔，在femtocell亦同樣會被F-GW阻隔，而影響用戶的連網成功率。 針對用戶需求特性，假設用戶使用數據服務的行為與進行語音服務的產生雷同，以Poisson隨機變數描述。連結流量分析模型採用何孟翰, 2011所提出的以語音服務為基礎；數據服務流量成分包括macrocell與femtocell子系統用戶的原始產生數據需求，以及因移動而由原子系統切換到另一子系統的數據需求流量，唯依據數據服務不同統計特性加以修改。數據連結流量具有不同的新進連結到達率，且總連結數據流量到達率(包含新進連結與切換連結)依何孟翰, 2011模型可以Poisson過程建模。在單一連結的特性方面，考慮用戶瀏覽網頁及觀看視訊服務，依據T. C. Wang et al., 2003，我們使用指數分佈模型來描述連結的使用時間。系統(基站)服務特性，WCDMA技術以單一頻帶利用多個碼同時傳輸多個連結訊務流量，核心網路採用分封交換，多工分享連網頻寬。連網瓶頸在無線接口的Uu介面WCDMA頻寬限制，使得各數據連結傳輸速率會因同時連結的數目而改變，此特性以Processor Sharing, PS模型描述。實務上在macrocell CAC對接入網路的數據連結個數有一上限，WCDMA技術可提供之最大傳輸速率2Mbps與macrocell單一連結基本頻寬需求之比值r，連結傳輸速率因連結數目而變，並有一上限值，我們以有界(bounded)的狀態相依服務速率(State Dependent Service Rate, SDSR)模型描述。同樣地，對femtocell子系統亦有一連結數目上限。依據上述用戶連結需求的產生與流量、WCDMA規約和頻寬共享之特性，採用M/ /1//r-PS模型。 本論文結合J. Beckers, et al., 2001及W. Cohen, 1979所分別提出的M/G/1/∞-PS穩態機率分析以及有限人數下(finite source population)M/Er/1//r-PS的穩態機率分析方法，加入bounded-SDSR模型，其中有界(bounded)用以描述UMTS傳輸速率有一上限為2Mbps，以分別求解macrocell與femtocell阻隔機率解析解。得到解析解。相較於M/Er/1//r-PS的解，因有界傳輸速率限制，每一連結狀態下穩態機率將上升，因而阻隔機率較SDSR將上升。 為使用所建模型及所得阻隔機率，依據實際上使用者行為設計當使用者長時間利用無線資源進行數據服務或使用者長時間在室內環境下使用數據服務等情境對於連結阻隔機率的影響。利用MATLAB軟體針對這些例子進行數值實驗，結果如下： (1) 增設femtocell可有效降低macrocell系統的阻隔機率，因為femtocell提供部分用戶流量需求的接取而能減少用戶對macrocell接取流量。Femtocell設置數目由5個增加至20個，macrocell阻隔機率下降23.5%。 (2) Macrocell的阻隔機率戶隨著總流量需求的上升而逐漸升高。當總流量的增加幅度由0.4至1連結/秒，阻隔機率上升幅度為21%。 (3) 用戶在室內之停留時間越久，佔用femtocell資源機率越高，可降低macrocell阻隔機率，例如用戶在室內停留時間平均值由1700s增加至2400秒，下降幅度為1.5%。 總結本論文的本研究之貢獻如下： 1. 探討共存系統數據需求特性以及服務端連結傳輸特性，用戶利用一段持續時間使用數據服務，研究顯示可以依連結持續時間描述； 2. 因分封交換特性，傳輸速率與連結數目成反比，我們以PS模型描述此特性； 3. 並經由本研究分析UMTS傳輸速率的限制，與一般SDSR模型不同，我們另加進有界(bounded)的SDSR模型進行分析，以實際針對UMTS系統下共存系統的阻隔機率進行解析解。

The femtocell and macrocell coexisting system is a trend for future mobile communications. In the thesis, we discuss the relationships among user data demand, the wireless band resource limitation and the quality of service of data service. Under the connection admission control (CAC) design, we shall model and analyze connection blocking probability of the coexistence system. The problem setting is that a wireless service provider (WSP) owns the band license for accessing the 3G mobile networks and operates macrocell and femtocell systems. Connection demand and traffic from macrocell users are multiplexed into the core network, in which traffic was aggregated into the Iu-ps. As for femtocell users, traffic was aggregated into Iuh interface. The Admission Control for Coexistence Coordinator (ACCC) is the central coordinator for CAC of the co-existence networks. CAC is implemented in the Radio Network Controller (RNC) in macrocell and the Femtocell GateWay (F-GW) in femtocell, which put limit on data service connection traffic and affects the connection access probability. From literature survey of the characteristics of data connection demands and traffic, user mobility and data connection demand for mobile services are similar to voice usage patterns. We thus follow the traffic modeling of by Ho, 2011. The traffic model consists of three parts, the original connection generation in macrocells and femtocells respectively and the handoff traffic from other cells. We further extend the modeling methods to capture data service characteristics. Data service types considered are web browsing and video streaming. We assume that original data connection demand generation in each cell is a Poisson process. The total connection demand generation with handoff traffic stays Poisson under Ho, 2011 model. The connection holding times are exponentially distributed according to Wang et al., 2003. User data traffic is multiplexed by using the Wide Code Division Multiple Access (WCDMA) technique and transmitted to the core network. So the service is packet-switched and access resources are shared among multiple connections, for which a Processer Sharing (PS) model is proposed. In practice, the number of access connections, r, has a limitation in macrocells and femtocells according to CAC mechanism. CAC limitation on number of connections is due to ratio of the WCDMA wireless bands limitation of Uu interface and connection bandwidth demand. As the connection data rate is dependent of the number of connections simultaneously accessing a Macro- or a Femto- cell. It is therefore modeled as State Dependent Service Rate (SDSR) of the cells. Furthermore, WCDMA data rate in UMTS has a maximum value, 2Mbps. We add an upper bound to the per-connection data access rate provided by a cell. In light of WCDMA protocol and the data traffic characteristics, we model the user aspects and the base station as an M/ /1//r-PS queueing model.

We combine and extend the analysis methods for M/G/1//∞-PS by Beckers, et al., 2001, and the finite population model of Cohen, 1979, to analyze the M/ /1//r-PS queueing model with a bounded-State Dependent Service Rate (b-SDSR) and derive macrocell and femtocell blocking probability. Comparing to analysis results of the M/Er/1//r-PS model, the blocking probabilities may increase because of the bounds on maximum service rate per connection and maximum number of connections. Scenarios are designed to study how user usage factors of connection holding times and indoor times impact on blocking probabilities. Numerical results were obtained by using the mathematical software MATLAB as follows: i. Deployment of femtocells can effectively reduce connection blocking probability of a macrocell, because some of the users can access femtocells, and hence the traffic in macrocell can be relieved. When total system arrival rate equals to 1 connection/sec and the number of femtocells increases from 5 to 20, the blocking probability of the macrocell decreases by 23.5%. ii. When total arrival rate to the system increases, the macrocell blocking probability increases. It can be seen that when the system traffic increases from 0.4 to 1 connection per second, the blocking probability increases by 21%. iii. Under the condition of fixed mean connection times of a user, if users stay indoor environments for longer time, the macrocell blocking probability decreases. For example, when mean indoor times changes from 1700 to 2400 seconds, the blocking probability decreases by 1.5%. In this thesis, we discuss and analyze the characteristics of data service, and the adoption of a queueing model for analytical results, the contributions are as follows: i. Discuss the limitation of WCDMA wireless band. We can grasp the relationship of the user demand traffic end, the data service holds for some times, which we denote it as a connection holding time; ii. Analyze the packet-switched characteristics, the data rate is inversely proportional to number of connections, which we adopt a PS model;

iii. UMTS data rate limitation makes the SDSR different from our design. We add the bounded-SDSR for blocking probability calculation.