Using multiversioning to improve performance without loss of consistency (Abstract Only)

Abstract

Various mechanisms, such as locking, time-stamped based algorithms, optimistic methods, versioning, have been used to support data consistency in transactions. Depending on the application's environment, InterBase offers the choices of two models for transaction handling, the consistency model, and the concurrency model. This talk presents a comprehensive overview of how Interbase's unique concurrency model, implemented with a multi-generational record structure, can improve global transaction throughput, by allowing a transaction to read a stable view of the database, without having to wait for simultaneous updates to release access to the data. A discussion of how this concurrency model handles the six classic problems of transaction management is presented Lost updates, Dirty reads, Cursor instability, Nonreproducible reads, Phantom records, Update side effects.

Efficient Support of XQuery Update Facility in XML Enabled RDBMS

“…However, even if it immediately changes the base data, it can use the database transaction log to provide read access to the data prior to the change. This is the concept of consistent read (CR) where data can be read at a particular snapshot version prior to the data change [25] [30]. The data layer uses database redo/undo transaction log to provide consistent data read at particular snapshot version.…”

Towards Business Intelligence over Unified Structured and Unstructured Data Using XML

“…With timestamp based data version technique, the database would essentially allow time traversal query where history could be queried. Consistent read is well-practiced in RDBMS (Bamford 1992), so is time traversal query capability (Gawlick 2004). …”