由於電腦圍棋的競賽中有時間上的限制,必須在有限的時間之內搜尋專家知識庫的棋譜資訊,因此如何利用盤面資訊找出好點的策略的即時性問題顯得相當重要。 為了在有限時間內,加速搜尋的時間,本論文建置分散式搜尋系統,其中這套系統採用Hadoop雲端平台的HDFS和MapReduce功能,來管理專家知識庫的檔案分配和進行分散式運算時的任務分派。在Hadoop搜尋系統方面,本論文提出Hadoop搜尋系統與Hadoop結合主從式搜尋系統,兩種設計方法。最後,在本文中,我們與單機搜尋,進行效能比較,來驗證所提出的兩種方法。實驗結果顯示,Hadoop結合主從式搜尋系統較適合圍棋棋譜搜尋的應用。
Since the Computer Go Game has time limit, we have to query the book information of the expert knowledge base within limited time, so how to find good point for the board information in real time problem is very important. In order to accelerate the speed of querying within limited time. In this thesis, we have developed the distributed query system.This system uses functions which are HDFS and MapReduce of Hadoop cloud platform to manage the file allocation of the expert knowledge base and task assignment for distributed computing. For Hadoop query system, in this thesis, we propose two designs: Hadoop query system and Hadoop query system combined with Client-Server model. Finally, in this thesis, we compare the performance of them with single query system to verify the two designs proposed. Experimental results show that Hadoop query system combined with Client-Server model is the more appropriate application for the Go book query.