個人通訊服務網路資源管理問題之研究

Abstract

隨著無線通訊技術的進步，個人通訊服務(personal communications services)的需求日益殷切，在個人通訊服務網路中，使用者接發行動電 話可不受時間、地點的限制任意通話。此一移動性大都由智慧型網路( Intelligent Networks)支援，而在此架構下，個人通訊服務資源可分成 無線資源、有線資源、以及支援行動管理之資源三大類。近年來已有很多 關於個人通訊服務行動管理方面的研究，然而卻少有關於資源管理方面的 研究，在本篇論文中，我們將探討個人通訊服務資源管理之相關問題，並 提出無線資源管理及支援行動管理之資源管理方面的解決方法。 此一研 究可分成下列幾個主題：(1) 無線資源管理：此一部份針對有限的無線頻 率做最有效的運用，提出了幾個拍賣式(auction-based)資源分配法，解 決了過去RAMA演算法中缺乏效率及不公平分配的問題。有別於過去的研究 ，我們從手機短小輕薄省電的考量出發，使得這些演算法均簡單極易實作 於手機中。此外、我亦提出一資料封包存取協定來確保其服務品質，實驗 結果證實此一協定適用於GSM的環境。(2)註冊區域規畫：當前定位策略的 標準如IS-41及GSM，均為二階層式定位策略，亦即以HLR/VLR分別記錄使 用者手機永久與暫時位址，藉以減少複雜度，是故VLR所涵蓋的範圍（稱 為註冊區域）的大小與幾何形狀，均會影響定位策略之效率。過去甚少關 於註冊區域規畫的研究，而這些研究亦未考慮soft constraints的問題且 僅可求得相對極值解(local optimum)。此一研究將註冊區域規畫轉換成 數值最佳化問題，並以基因演算法求解，實驗結果證實此一模型更具一般 性、此一方法亦可以獲得極佳的註冊區域規畫結果。(3)位置資料庫容錯 ：由於使用者之位置資料均存於位置資料庫，一旦資料庫毀損則將無法打 電話至這些行動手機，而造成極大的損失。在此論文中我提出一個積極的 錯誤回復演算法，此方法採用不定時檢查點（aperiodic checkpointing ）方法，同時我們也比較定時檢查點（periodic checkpointing）與不定 時檢查點方法效能之優缺點，數值結果說明不定時檢查點將可使錯誤回復 時須傳送之資料減至最小。 With recent advances in wireless communications, we have seen a rapidly growing interest in Personal Communications Services (PCS). PCS networks enable subscribers to communicate with each other at any timefrom any location. The user mobility is generally supported by Intelligent Networks (IN). According to the IN-based PCS reference model,PCS resources can be classified into wireless resources, wireline resources, and resources for mobility management. In recent years, many researchers have contributed to mobility management ofPCS networks.However, there has been little research on PCS resource managementin the literature. In this dissertation, we concentrate on PCS resource management and contribute our solutions to some critical issues on wireless resource managementand resource management for mobility support. For wireless resource management, we have proposed several auction-based protocols for fair andfast resource assignment. The proposed protocols resolve the unfairness problem and adopt dynamic priority assignment to improve the QoS for subscribers in overload environments. Moreover, we also propose a deterministic packet access protocol, called Q-RAMA, for QoS-guaranteed services. Simulation results show that the Q-RAMA protocol achieves better performance in terms of low packet dropping probabilities and low speech clipping frequencies in GSM-like environments. For mobility-support resource management, we have proposed a genetic approach to the problem of registration area planning. In this approach, the problem of registration area planning is formulated as a cost optimization problem. This model is general and suitable for soft presets and constraints. Moreover, performance results also show that the proposed genetic approach isrobust in avoiding getting trapped into local minimums for registration area planning. In addition, we also study the issue of mobility database recovery which is concerned with how to locate a subscriber after mobility database failure. The proposed restoration algorithm is based on an aperiodic checkpointing scheme to back up VLRs and HLRs. Performance analysis shows that aperiodic checkpointing is more feasible than periodic checkpointing for aggressive restoration.