Altruistic object versioning scheme for active data warehousing

Abstract

In order to speed up on-line analytical processing (OLAP), data warehouse, which is usually derived from operational databases, has been introduced. When operational databases happen to change, data warehouse gets stale. To enhance the freshness of data warehouse, operational databases`` changes need to be frequently and concurrently propagated into the data warehouse. However, if several update transactions are allowed to execute concurrently without an appropriate concurrency control, data inconsistency between data warehouse and operational databases can arise due to incorrect propagation of changes on the operational databases into the data warehouse. In this thesis, we propose a new concurrency control scheme, called altruistic object versioning scheme (AV), which is able to execute a number of update transactions in a concurrent and consistent way. To enable long-lived OLAP transactions to read fresh data at the expense of relatively low I/O overheads, AV attempts to reduce the number of versions maintained. Instead of creating a new version of a data whenever the data is updated, an update transaction is allowed to create a new version of the data in a selective way only when it makes any conflicts against other concurrent OLAP transactions. Moreover, to minimize an additional disk space required for version management, AV also attempts to exploit the recovery log file as a transient version pool. Owing to this selective versioning with recovery log, AV will be able to reduce the number of versions maintained as well as disk space for version management. To investigate the applicable areas of AV, its performances have been evaluated by means of simulation approach. Our experimental results show that AV is capable of providing a reasonable performance under various workloads.