以快取服務提升電子商務網站收益之研究

過去快取置換機制以提高快取的效能為主，往往僅考慮文件上次的擷取時間、過去擷取的次數、或是文件大小來置換快取內的文件內容。但是從網站經營的角度來思考，使用者能快速存取的資訊，最好是較能促進其消費行為，俾使網站獲得更大的利益。為達此目的，本研究建構了一個考慮文件（資訊）價值的快取置換機制。此機制考慮四個參數：文件上次擷取時間、文件擷取次數、文件大小、以及文件的主觀資訊價值以決定文件置換之優先順序。為求取文件的資訊價值，我們應用網路探勘的技術，分析使用者的瀏覽路徑中，各網頁與目標網頁的關聯強度，最後根據一公式決定出各個網頁的資訊價值。當快取有執行置換之須要時，本置換機制將優先換掉資訊價值低、上次擷取時間久、擷取次數少、或太大的文件。根據初步之實驗結果，其運作效率較傳統的四種置換機制為佳，特別是在消費者佔總瀏覽人數的比例較高的時候。

關鍵字

資訊價值；快取置換機制

並列摘要

Document caching service is usually provided by an EC website to lower the burden of the system server and to promote the quality of service of the website. However, limited to the capacity of the cache server, not all documents of the website can be simultaneously kept on the cache server. As a result, ”cache miss” happens when a document requested by a customer does not exist on the cache server. A good cache control mechanism can significantly reduce the probability of cache miss. Among all the factors influencing the performance of caching service, document replacement policy is undoubtedly the most critical. There have been many researches devoted to developing efficient cache document replacement policies, but most of them did not consider whether the cached information was value-added to the EC website or not. As a result, the cache server might keep responding to the requests of pure browsers instead of fulfilling the needs of customers that will purchase in the website. To solve this problem, a new cache document replacement policy considering the values of web pages is proposed in this study. A web page is considered to be more valuable if it's associated with a purchasing behavior. To evaluate the performance of the proposed cache document replacement policy, an experimental system has been constructed in this study.

並列關鍵字

Information Value ； Cache Replacement Policy ； EC

參考文獻

Arlitt, M.,Cherkasova, L.,Dilley, J.,Fridrich, R.,Jin, T.(1999).Proceedings of the 2(superscript nd) Workshop on Internet Server Performance.Atlanta GA:

Google Scholar

Arlitt, M.,Krishnamurthy, D.,Rolia, J.(2001).Characterizing the scalability of a large web-based shopping system.ACM Transactions on Internet Technology.1(1),44-69.