適用於關聯式線上分析處理資料的索引架構之設計與研究

傳送到

題名:
適用於關聯式線上分析處理資料的索引架構之設計與研究
著者: 陳科學
李建億; 資訊教育研究所
主題: 總集查詢; 高維度環境索引
描述: 摘　要資料倉儲是結合資料庫與決策支援的產物，作為資料庫的後端作業。儲存於資料倉儲中各種巨細糜遺的資料，有助於提供線上分析處理(On Line Analysis Processing，OLAP)，允許透過資料庫查詢語言(SQL)，提供即時的整合性資訊。OLAP依照實作方法，又可區分為關聯式線上分析處理(Relational On Line Analyze Processing，ROLAP)和多維資料線上分析處理(Multi-dimension On Line Analyze Processing，MOLAP)兩種。ROLAP適用在資料密集度較低、分布較不均衡的環境下運作，並以樹狀資料結構為索引；本研究的重點則是以ROLAP作為探討的對象。線上分析處理的目的是提供快速且整合性的資料，幫助使用者以最有效率的方式，觀察趨勢、判斷機會、進行分析和預測未來。資料方格(Data Cube)則是描述的OLAP中多維屬性聚集化(Aggregation)架構的模型。在實務上，ROLAP必需符合快速更新、即時反應的要求，且必需兼顧記憶儲存空間的限制。經文獻研究發現：R-tree是最適合於ROLAP的索引結構。其中又以於R*-tree目錄層存放子樹特徵值，並改寫總集查詢方法的R*a-tree為最。然而經由我們研究發現：R-tree及其相關方法的樹狀結構，其目錄節點在高維度環境下嚴重的overlap，使得無效搜尋次數增加，影響查詢效率。因X-tree有著接近於R*-tree的建構方法，且改良R*-tree於中高維度時嚴重overlap的缺點，我們針對ROLAP環境需求，改良X-tree，結合R*a-tree於非葉節點存放子節點的特徵值的特性，獲取更佳的總集查詢效率。實驗結果證明：改良自X-tree的Xa-tree，在中維度空間資料範圍的總集查詢，擁有高於R*a-tree的效率，比R*a-tree更適於中維度空間的資料範圍的總集查詢，可提高ROLAP的索引效能。
Abstract Data warehouse is the product which combined by database and decision supporting system. Being the tail task of database system, the data stored in data warehouse help to support On Line Analysis Processing(OLAP). It permit to access by SQL, provide integrate information at once. According to the way of implementation, OLAP can be classfied into ROLAP (Relational On Line Analysis) and MOLAP (Multi-dimensional On Line Analysis). In general, ROLAP is suitable for the cases which data density is low, or data distribution is uneven. This thesis is focus on ROLAP research. The goal of ROLAP is to provide integrate data on the moment, help user to find the future trend, decision making, data analyzing, and forecast its opportunity in the most efficient way. Data cube is the model to describe multi-dimensional data attributes in ROLAP. In practice, ROLAP must conform to fast update, run in real times, and look after for the restriction of storage facilities. From the study of relative work, we found that R-tree-like indexes(especial, R*a-tree)are mostly suitable for a ROLAP application. However, all the R-tree like index methods have one major drawback, called overlap, results in some unnecessary search paths and degenerates performance in high dimensional data space .On the other hand, X-tree , which is the similar method index to R*-tree, has been proved to present higher performance in multi-dimensional space. Therefore, as the modified version R*a-tree from R*-tree, we expend X-tree by adding an aggregate feature value in each non-leaf node to accelerate aggregate query, which is called Xa-tree. From our experiment, we find that Xa-tree has better aggregate query performance in medium dimension then X-tree , R*-tree , and R*a-tree .
碩士
建立日期: 2000
格式: 121 bytes
text/html
語言: 中文
識別號: http://nutnr.lib.nutn.edu.tw/handle/987654321/4806
資源來源: NUTN IR

連結

連回原始介面

適用於關聯式線上分析處理資料的索引架構之設計與研究

陳科學李建億; 資訊教育研究所 2000

正在檢索遠程資料庫，請稍等

適用於關聯式線上分析處理資料的索引架構之設計與研究

陳科學 李建億; 資訊教育研究所 2000

正在檢索遠程資料庫，請稍等

陳科學李建億; 資訊教育研究所 2000