Spatial OLAP: Open Issues and a Web Based Prototype Sandro Bimonte, Anne Tchounikine, Maryvonne Miquel LIRIS (Laboratoire d'InfoRmatique en Images et Systèmes d'information) UMR CNRS 5205 INSA, 7 Avenue Capelle, 69621 Villeurbanne Cedex, France Name.Surname@insa-lyon.fr INTRODUCTION A Data Warehouse (DW) is a centralized repository of data acquired from external data sources and organized following a multidimensional model (Inmon, 1996) in order to be analyzed by On-Line Analytical Processing (OLAP) applications. OLAP tools provide the ability to interactively explore multidimensional data presenting detailed and aggregated data. The results of the analyses are the basis of strategic business decisions. Spatial information is very often embedded in data but despite its significance, multidimensional models usually treat them as textual dimension. Integration of valuable spatial data in DWs leads to the definition of Spatial OLAP (SOLAP) (Rivest, 2005). SOLAP applications can address several and different domains: environmental studies, marketing, archaeology, epidemiology, etc… Introduction of spatial data in multidimensional model raises major problems from implementation and theoretical point of view. This paper makes an inventory of some open issues in SOLAP. Then we propose a web based GIS-OLAP integrated solution supporting geographical dimensions and measures, and providing interactive and synchronized maps, pivot tables and diagrams displays in order to effectively support decision makers. OLAP AND DATAWREHOUSING In multidimensional models, data is organized as an n-dimensional cube or hypercube. Figure 1 shows an example of a three-dimensional cube, which represents the sale of products in cities within a given time. Each dimension, i.e. cube’s axes, can present a hierarchical structure which organizes data at different levels of details usually based on classification or generalization/specialization relationships. Facts are described by numerical values, called measures and stored in the cube's cells. (e.g. the number of items sold). Typical OLAP queries can involve multiple selections, nested group- bys and aggregated values, e.g. “Display the number of TVs sold per cities in USA and per month, and display the subtotals for each quarter”. The result is a multidimensional and multi-levels table. Data warehousing architectures are usually three-tier architecture (figure 2). The first tier is a warehouse server where data of interest is loaded after being extracted, cleaned, transformed and homogenized from operational legacy databases. Figure 1: Sales Cube. 10th AGILE International Conference on Geographic Information Science 2007 Aalborg University, Denmark Page 1 of 11