通訊網路障礙識別方法之研究

Abstract

複雜之通訊網路與其所傳輸之大量資訊，已使網路管理工作漸趨困難。因障礙是無可避免，為使得網路系統更為堅固耐用且其操作更為可靠，系統具備快速偵測、識別與回復能力是非常重要的。本論文就通訊網路之障礙識別作業提出四種事件關連方法；其中，因果圖模式(causal graph model)是用來描述網路事件間之因果關係。所提之兩種以編碼法為主(coding-based)之方法，係植基於Yemini等人所提之編碼法，對每個問題及每個癥狀會賦予一個二元碼，而針對癥狀碼使用Boolean運算將使得關連過程甚為快速。以集合為主(set-based)之方法則是植基於集合之代數運算，其中每個問題及每個癥狀會賦予一個質數，而利用最大公因數(GCD)使得關連過程簡單且快速。以排程為主(scheduling-based)之方法是植基於雙層回饋列(bi-level feedback queues)排程策略，而該策略會使得關連程序非常快速。以模擬方式驗證所提方法之效益與效率。從模擬結果得知，所提之方法不僅能一次識別多個問題並且不易受雜訊所影響，且所提方法之時間複雜度均比Yemini之編碼法來得低。

The complexity of communication networks and the amount of information transferred in these networks have made the management of such networks increasingly difficult. Since faults are inevitable, quick detection, identification, and recovery are crucial to make the systems more robust and their operation more reliable. This dissertation proposes four event correlation schemes for fault identification in communication networks. The causality graph model is used to describe the cause-and-effect relationships between network events. Two coding-based schemes are based on the coding scheme proposed by Yemini et al. For each problem, and each symptom, a unique binary code is assigned. The use of Boolean operations on symptom code makes the correlation process very fast. The set-based scheme is based on the algebraic operations of sets. A unique prime number is assigned to each problem, and each symptom. The use of the greatest common devisor (GCD) makes the correlation process simple and fast. The scheduling-based scheme is based on the bi-level feedback queues scheduling policy. The scheduling policy makes the correlation process very fast. Simulation models are developed to verify the effectiveness and efficiency of the proposed schemes. From simulation results, we notice that all proposed schemes not only identify multiple problems at one time but also are sensitive to noise. For all schemes, their time complexities are much lower than the time complexity of the original coding scheme.