Abstract
This paper is concerned with efficient querying of very large multi-resolution datasets on storage and compute clusters. We present a suite of services that support storage, indexing, and data processing (data sampling and data aggregation) on datasets that consist of a collection of multi-resolution Grids. We empirically evaluate the performance impact of different data declustering, indexing, and query processing strategies. The experimental evaluation is carried out using a data server implemented to serve multi-terabyte multi-resolution volumetric datasets to remote visualization clients and a one-terabyte multi-resolution volumetric dataset on a PC cluster with distributed disk space.