目前，在地球科學、地理、天文等領域的研究或應用中，需要分析處理空間中的物件，而物件的位置通常是在球面上以赤經及赤緯來記錄。為了使儲存及管理這些物件更有效率，需設計出一套良好的球面空間索引系統。隨著科技的進步以及資料長久的累積，各個領域需要分析的資料量越來越多，導致在傳統單機應用的環境下，硬體無法負擔。
因此，本論文建構了分散式球面空間索引系統，以多層次三角網格（HTM）為模型，運用Hadoop的MapReduce運算框架，設計出相應的資料結構及分散式演算法。不但能儲存及管理大量球面空間物件，還能提供分散式應用程式快速索引與查詢的功能。
本論文之分散式球面空間索引系統，是使用泛星巡天計劃Desktop Virtual Observatory (DVO) 資料集，而經由實驗將系統中各項參數對效率的響影做比較，供使用者參考，可以針對不同的系統環境及不同的應用需求進行調整，達到系統的最佳效率。

Recently, research or applications in earth science, geography, astronomy, those need analyzing and processing spatial objects which position is usually recorded by right ascension (ra) and declination (dec) on sphere. In order to store and manage those objects efficiently, we need to design a well sphere spatial index system. With the progress of technology and long-term accumulation of data, there are more and more data need to analyze in each field, this result cause that hardware can’t afford data in traditional single machine application environment.
Therefore, this paper construct distributed sphere spatial index system which is modeling by Hierarchical Triangular Mesh (HTM) using Hadoop MapReduce computing framework to design reference data structure and algorithm. Not only can it store and manage lots of sphere spatial objects, but offer distributed applications indexing efficiently and searching function.
Distributed sphere spatial index system in this paper, using Desktop Virtual Observatory (DVO) data in Pan-STARRS Project. Through experiments, we compare influence of efficiency by each parameters in this system, and that results can be referenced by users. They can modify parameters according to different system environments and different applications, and distributed sphere spatial index system can achieve its’ best efficiency.

摘 要 i
Abstract ii
致 謝 iv
目 錄 v
圖目錄 vii
一、緒論 1
1-1 研究動機 1
1-2 研究目的 2
1-3 論文章節介紹 3
二、文獻探討 4
2-1 泛星計畫 4
2-2 OpenStack 4
2-3 Hadoop 5
2-4 多層次三角網格 7
三、系統架構與流程 15
3-1 系統架構 15
3-2 系統流程 15
四、研究方法 17
4-1 HTM索引樹 17
4-2 CreateIndex：建構HTM索引樹之MapReduce任務 20
4-3 PointQuery：點查詢之MapReduce任務 23
4-4 RangeQuery：範圍查詢之MapReduce任務 24
4-5 單機訪問HDFS之查詢任務 28
五、實驗 29
5-1 實驗環境與資料集 29
5-2 CreateIndex執行時間 30
5-3 PointQuery執行時間 31
5-4 RangeQuery執行時間 35
5-5 樹根深度影響 38
5-6 葉節點容器值影響 39
5-7 集中資料集查詢 41
六、結論 44
七、參考文獻 45

[1] SkyServer – HTM, http://www.skyserver.org/
[2] Pan-STARRS, http://pan-starrs.ifa.hawaii.edu/public/
[3] TAOS, http://taos.asiaa.sinica.edu.tw/
[4] SUBARU, http://www.subarutelescope.org/index.html
[5] Apache Hadoop, http://hadoop.apache.org/
[6] OpenStack, https://www.openstack.org/
[7] Sachin Puttur: Big Data: Overview of apache Hadoop, http://www.sachinpbuzz.com/2014/01/big-data-overview-of-apache-hadoop.html
[8] Hadoop 101: Programming MapReduce with Native Libraries, Hive, Pig, and Cascading, http://blog.pivotal.io/pivotal/products/hadoop-101-programming-mapreduce-with-native-libraries-hive-pig-and-cascading
[9] A. Szalay, J. Gray, G. Fekete, P. Kunszt, P. Kukol, and A. Thakar, "Indexing the sphere with the hierarchical triangular mesh", in MSR-TR- 2005-123, September 2005.
[10] Feng, Jun, et al. "HQ-Tree: A distributed spatial index based on Hadoop." China Communications 11.7 (2014): 128-141.
[11] Lv, Zhenhua, et al. "Spatial indexing of global geographical data with HTM." 2010 18th International Conference on Geoinformatics. IEEE, 2010.
[12] 鄭凱豪, "Distributed Large-Scale Astronomical Data Management System Based on HTM Index", 國立中央大學, 碩士論文, 2012.
[13] 黃郁哲, "Hierarchical Distributed Index and File System for Astronomical Observation Data", 國立中央大學, 碩士論文, 2014.