本論文在LISP 協定下，提出了基於貝氏網路之快取生命週期決定方法（BNCLD）來計算合適的cache lifetime方法。貝氏網路在處理具有不確定性的相關知識時具有相當優勢，因為貝氏網路可以經由觀測到的證據或已知的背景知識對未知或具有不確定性的狀態進行推論。BNCLD透過資料萃取模組抓取可以從網路上觀測到的資料，並將觀測資料交由運算模組。以貝氏網路為核心進行運算，最後推得出合適的cache lifetime。模擬結果顯示BNCLD 的效能比Fixed Time Method（FTM）優異。BNCLD 在average ITR delay較之使用FTM減少可達32.57%，packet loss ratio減少可達35.24%，control overhead上減少可達33.58%。;Due to the usage of IP addresses rises rapidly, researchers proposed Locator/ID Separation Protocol (LISP) to solve the routing scalability problem. LISP reserves the “identifier” function of IP address. LISP uses Ingress Tunnel Router (ITR) and Egress Tunnel Router (ETR) to represent the “locator” function. LISP uses mapping system to map identifier and locator. To ease the heavy burden of mapping system, ITR has the cache equipment. The value of cache lifetime will influence the times that ITR queries mapping system. Therefore, deciding an appropriate cache lifetime can reduce the load of mapping system.
A Bayesian Network Based Cache Lifetime Determination scheme (BNCLD) for LISP is proposed in this thesis. According to the observed quantities or domain knowledge, Bayesian network can solve problems under uncertainty. BNCLD uses the Information Retrieve Module (IRM) to collect parameters from the Internet. BNCLD uses Computation Module (CM) which is based on Bayesian network to infer an appropriate cache lifetime. According to simulation experimental results, BNCLD outperforms than Fixed Time Method (FTM). BNCLD has shorter average ITR delay than FTM up to 32.57%. BNCLD improves the packet loss ratio up to 35.24%. BNCLD can reduce the control overhead up to 33.58%.