基于HBase的瓦片索引和读取方法设计

DESIGN OF TILE INDEX AND QUERY METHOD DESIGN BASED ON HBASE

地理信息瓦片数据规模庞大,传统的关系型数据库不足以满足对瓦片数据容量、速度的支持。针对这种情况,选取HBase存储瓦片数据,在以四叉树编码作为索引进行数据存储的基础上,提出一种多层级瓦片数据索引设计方法,该算法通过“异层级再编码”的方式,使在地图中空间相邻的瓦片数据在物理存储中也尽可能相邻,以达到瓦片数据的空间连续性在物理存储中得以保持的目的;提出一种邻域预读取方法,该方法配合索引设计,以批量读取的方式预先将邻域瓦片读取至缓存。实验表明,在用户连续读取相关联的地图数据的情况下,该方法比传统的缓存算法具有更高的命中率,且比传统的数据索引设计响应更快。

Geographic information tile data is so large that traditional relational databases are not sufficient to support the capacity,speed and type of tile data.In this case,HBase is selected to store tile data.On the basis of quadtree encoding as an index for data storage,a multilevel tile data index design method is proposed.The algorithm made the tile data adjacent to each other in the map as adjacent as possible in physical storage to achieve the purpose of maintaining the spatial continuity of tile data in physical storage.A neighborhood pre-reading method was proposed,which worked with index design to pre-read neighborhood tiles into the cache based on batch reading.Experiments show that this method has a higher hit ratio and faster response than traditional data index design when users continuously read the associated map data.