Clustering of Shared Subobjects in Databases.
Jukka Teuhola · 1993
The topic of this article is multi-criterion, structure-based clustering in objectoriented databases. We study an object class, which is the target (subobject) of several multi-valued reference types from other object classes. The aim is to serve all access paths fairly, so that the number of page accesses is proportional to the number of referenced occurrences of the subobject class. An efficient heuristic algorithm is developed for inserting new subobjects in an existing page set. Significant benefits can be obtained in read-intensive applications, if the confluent references are semantically correlated. Keywords: Clustering, Page allocation, Object-oriented databases 1 Introduction Clustering of objects (records) is necessary for effective database processing, as long as they are stored on conventional disks with slow random access. Three general kinds of clustering can be distinguished: 1. Content-based clustering: Objects sharing the same value for a certain attribute are placed...