A performance evaluation of g-bound with a consistency protocol supporting multiple isolation levels

1Citations
Citations of this article
4Readers
Mendeley users who have this article in their library.
Get full text

Abstract

Whereas the Strict Snapshot Isolation (SSI) level is nowadays offered by most of the centralized DBMSs, replicated databases do not usually provide it, since the traditionally considered approach, the pessimistic management, introduces enormous performance penalties that preclude it from production applications. Instead, distributed databases usually offer Generalized Snapshot Isolation (GSI), a more relaxed level, in which a transaction may get a snapshot older than the one that was applied on the database by the time the transaction started. This paper takes advantage of our MADIS middleware and one of its implemented Snapshot Isolation protocols (SIRC) to design, implement and evaluate the performance of an extended version of SIRC (called gBSIRC). This protocol is able to concurrently execute Generalized Read Committed (GRC), GSI, g-Bound —a non-standard SI level limiting the outdatedness of transactions wanting to commit— and optimistic SSI transactions on top of a cluster of centralized DBMSs offering RC and SSI. This work is the first implementation and evaluation of an optimistic SSI level. Although the abort rate of g-Bounded transactions is significantly higher than the GSI ones, the performance results show that introducing transactions at more restrictive levels is not detrimental to the completion time or to the abort rate of the transactions using GSI.

Cite

CITATION STYLE

APA

Salinas, R., Muñoz-Escoí, F. D., Armendáriz-Iñigo, J. E., & González de Mendívil, J. R. (2008). A performance evaluation of g-bound with a consistency protocol supporting multiple isolation levels. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5333, pp. 914–923). Springer Verlag. https://doi.org/10.1007/978-3-540-88875-8_117

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Save time finding and organizing research with Mendeley

Sign up for free