R-trees Have Grown Everywhere

Spatial data management has been an active area of intensive research for more than two decades. In order to support spatial objects in a database system several issues should be taken into consideration such as: spatial data models, indexing mechanisms, efficient query processing, cost models. One of the most influential access methods in the area is the R-tree structure proposed by Guttman in 1984 as an effective and efficient solution to index rectangular objects in VLSI design applications. Since then, several variations of the original structure have been proposed towards: providing more efficient access, handling objects in high-dimensional spaces, supporting concurrent accesses, supporting I/O and CPU parallelism, efficient bulk-loading. It seems that due to the modern demanding applications and after the academia has paved the way, recently the industry recognized the use and necessity of R-trees. The simplicity of the structure and its resemblance to the B-tree, allowed developers to easily incorporate the structure into existing database management systems in order to support spatial query processing. In this paper we provide an extensive survey of the R-tree evolution, studying the applicability of the structure and its variations to efficient query processing, accurate proposed cost models, and implementation issues like concurrency control and parallelism. Based on the observation that ‘space is everywhere’, we anticipate that we are in the beginning of the era of the ‘ubiquitous R-tree’ in an analogous manner as B-trees were considered 25 years ago.