关键词: 聚集计算, 近似查询, 空间索引, 在线分析

Abstract: While coping with the soaring spatial data, spatial aggregation proves to be competent and efficient, though it can be compute-intensive. In terms of spatial online aggregation, traditional stand-alone serial methods gradually become limited. However, the current parallel computing architectures widely used nowadays, scarcely have research conducted on the index-based parallel online aggregation methods specifically for spatial data. Therefore, two new indexes-based methods are proposed to support spatial online aggregation analysis. In the first method,     indexes are organized in two-layers, where the global grid index filters the related local indexes and the local indexes accelerate the aggregate query locally. In the second method, on the basis of the first method, the random sampling, adaptive data-bricks partition, dynamic caching, and other optimization techniques are all applied. In this way, when given certain confidence, the final results are returned with certain credit intervals. Experimental and analytical results on billion-scale data verify the effectiveness and scalability of those methods.

Key words: aggregation computation, approximate query, spatial index, online analysis