A modular history-oriented access structure for bitemporal relational databases

Abstract

Two time dimensions are usually considered in temporal databases: transaction-time, which tells when an event is recorded in a database, and valid-time, which tells when an event occurs, occurred or is expected to occur in the real world. Transaction-time is defined by the system and can only grow, since it records the time of successive transactions. This peculiarity makes it difficult to adapt traditional access structures for non-temporal data, such as indices or grid files, since transaction-time unbalances such structures. The maintenance of access structures for temporal data can thus be very expensive, also due to the large amount of data contained in temporal relations. In this paper, we propose a variant of an index composed by as many sub-indices as the number of objects. The main feature of this structure is to localize the unbalancements and thus, whenever a sub-index needs reorganization, not to burden the whole structure.