A Map-Reduce-enabled SOLAP cube for large-scale remotely sensed data aggregation

Abstract

Spatial On-Line Analytical Processing (SOLAP) is a powerful decision support systems tool for exploring the multidimensional perspective of spatial data. In recent years, remotely sensed data have been integrated into SOLAP cubes, and this improvement has advantages in spatio-temporal analysis for environment monitoring. However, the performance of aggregations in SOLAP still faces a considerable challenge from the large-scale dataset generated by Earth observation. From the perspective of data parallelism, a tile-based SOLAP cube model, the so-called Tile Cube, is presented in this paper. The novel model implements Roll-Up/Drill-Across operations in the SOLAP environment based on Map-Reduce, a popular data-intensive computing paradigm, and improves the throughput and scalability of raster aggregation. Therefore, the long time-series, wide-range and multi-view analysis of remotely sensed data can be processed in a short time. The Tile Cube prototype was built on Hadoop/Hbase, and drought monitoring is used as an example to illustrate the aggregations in the model. The performance testing indicated the model can be scaled along with both the data growth and node growth. It is applicable and natural to integrate the SOLAP cube with Map-Reduce. Factors that influence the performance are also discussed, and the balance of them will be considered in future works to make full use of data locality for model optimisation.