應用於雲端服務之內容分發網路(Content Delivery Network ; CDN)是一個利用預載、 分配副本在全球各個節點的計算機系統。CDN 如果得到正確的實施，可以提高網站入口 的頻寬利用並降低Query 的延遲時間。常在CDN 快取的資料包含網頁資料 (如文件、圖 片、網址) 、下載的資料 (如影片、軟體、文件)，以及應用程序、實況直播、資料庫搜 尋等。CDN 技術可以分成兩個類型︰(1)邊緣伺服器的選擇、(2)快取儲存。邊緣伺服器 的作用，就是將以往Query 遠距離傳送到伺服器以取得服務，改為利用距使用者最近的 邊緣伺服器來服務這筆Query，所謂距使用者最近的邊緣伺服器，即是由CDN 的DNS (Domain Name System) 計算出服務該Query 最佳的邊緣伺服器。第二類是快取的建立、 分配和維護，包括邊緣伺服器裡所有資料的維護、更新及取代機制，而邊緣伺服器僅會 接受由CDN 所提供的資料。 在這個計畫中，我們將建構一個模擬環境並利用這個環境來測試任何情況下，現今 各主要CDN 快取機制的效能高低。此外，我們也將建立一套計算成本的機制，用以評 估計算每個 既有CDN 快取機制的複雜度。最後我們將歸納每個 既有CDN 快取機制的命 中率、失誤率、快取正確率、安全性和計算複雜度，據以分析瞭解其優缺點，進而在適 當合理的計算複雜度下，開發出一個更有效率的CDN 快取機制，進一步強化入口網站 的使用。 A content delivery network (CDN) in cloud services is a computer system containing copies of data preloaded and distributed at nodes around the world. If properly designed and implemented, a CDN can improve data access efficiency by increasing bandwidth utilization and reducing access latency. Data that are frequently cached in CDNs include (1) web objects – such as text, graphics, URLs and scripts, (2) downloadable objects – such as media files, software and documents, (3) applications, (4) live streaming media and (5) database queries. A CDN has two parts: edge server selection and caching. In traditional networks, a client needs to uplink a query to the server in order to access the required data; this tends to cause bandwidth waste and access delay when the client is far away from the server. For improvement, a CDN can follow the obtained DNS’s (Domain Name System’s) calculation results and pick up a nearest edge server – a single or group computer located near the client – to provide the client with the requested cached item. A replication server is used to transfer such distant requests to the nearest suitable edge server. Its function includes building and maintaining cached data items stored in the edge servers which only backup the data objects distributed from the CDN. In this research, we plan to construct a CDN simulation environment and implement existing CDN caching mechanisms in this simulated environment to evaluate their performance and to locate their advantages as well as disadvantages. We will further analyze their hit ratio, miss ratio, cache hit ratio, security and computational complexity by building a proper computational evaluation module. Our ultimate goal is to develop an efficient new CDN caching mechanism with reasonable computation complexity. By preserving the advantages and avoiding the disadvantages of previous mechanisms, we expect to come up with a more advanced CDN caching mechanism which is able to cut down the needed bandwidth consumption and data access time at reasonable cost.