An Improvement to the Length Estimation Based Methods to Analyse the Performance of R*-Trees

R* -trees were motivated by the need to efficiently process window queries in a multidimensional space. Many indexing methods were proposed to index the spatial data efficiently. All these methods use several heuristics to index spatial data efficiently. The performance efficiency of a spatial data indexing structure is usually evaluated by the ability to answer queries by accessing the smallest possible number of pages in the disk. Today, research in this field has turned towards the development of powerful analytical methods to predict the performance of the spatial data indexing structures for arbitrary inputs. While several diverging approaches are made to address this issue, approaches based on estimating the MBR extents have gained momentum and are applicable to generic data rather than data belonging to specific domains. This paper proposes improvements to those methods that estimate the length of extents of MBRs to predict the number of disk access that are required to answer a window query. The amendments significantly improve the performances.