Slashgeo

The OSGeo-Discuss mailing list links to a short survey of geospatial databases such as MySQL, Oracle, IBM DB2 UDB and PostGIS. From the accompanying article: "For example, subjects like ‘spatial indexing’, ‘optimising approaches’, ‘spatial joining algorithms’ and ‘topology management’ may relate to very sophisticated approaches not easily condensed and described in a survey matrix box. In my first Product Survey on RDBMS in 2002 I observed that spatial-data management was traditionally supported within GIS environments. Integration with other datasets has to be organised within such environments with the aid of complementary architectures. As more and more support for spatial-data management becomes available in RDBMS, direct integration with other datasets may be organised in an increasingly flexible way. This allows for fast data access, easy product development, avoidance of duplication of spatial data etc." See numerous related articles below.

Using pyramids for raster geodata can radically improve load speed. All Points Blog offers several informative comments on raster data pyramids. From comments: "Databases are not inherently more efficient than files, particularly when the files in question, images, have a great deal of internal structure. You don't need a spatial index to do random access of small chunks of large images, the "square grid" nature of the image itself provides a built-in indexing mechanism. [...] You must choose between the options carefully : pre-calculating avoids completely to set up a application on your server, since you just have to distribute resulting files. This has a huge impact on server deployment costs, maintenance and hardware requirements. The downside is that generating the whole pyramid is a big process and let you to manipulate an horrible amount of files." See also related stories.

The Lin.ear th.inking blog ran an entry last month on spherical indexing schemes for PostGIS. Thought not exhaustive, the entry is still interesting and reminds us how spatial indexes are important to efficiently analyze huge spatial databases. From the entry: "Handling geodetic data in a correct and efficient way presents quite a few challenges. A major one is: how can geodetic geometry be spatially indexed? Conventional spatial indexes (such as 2D R-trees) all rely on geometry being embedded in a planar space. They don't handle data which can "wrap around", as can occur in a spherical space." See below for other (rare) stories related to spatial indexes.