This paper concerns the problem of clustering objects onto units of secondary storage to minimise the number of I/O operations in database applications. We first investigate problems associated with most existing clustering schemes. We then propose STD, a Statistic-based Tunable and Dynamic clustering strategy which is able to overcome deficiencies of existing solutions. Our main contributions concern the dynamicity of the solution without adding high overhead and excessive volume of statistics. Reorganisations are performed only when the corresponding overhead is strictly justified. Clustering specifications are built from observation upon objects life, capturing any type of logical or structural inter-object links. Moreover, our clustering mechanism does not need any user or administrators hints, but remains user-controlled. A partial validation of STD has been made using Texas.
CITATION STYLE
Bullat, F., & Schneider, M. (1996). Dynamic clustering in object databases exploiting effective use of relationships between objects. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1098, pp. 344–365). Springer Verlag. https://doi.org/10.1007/bfb0053069
Mendeley helps you to discover research relevant for your work.