Snapshot isolation (SI) is a widely studied concurrency control approach, with great impact in practice within platforms such as Oracle or SQL Server. Berenson et al. showed though that SI does not guarantee serializable execution; in certain situations, data consistency can be violated through concurrency between correct applications. Recently, variants of SI have been proposed, that keep the key properties such as (often) allowing concurrency between reads and updates, and that also guarantee that every execution will be serializable. We have had the opportunity to use three implementations of two different algorithms of this type, all based on the InnoDB open source infrastructure. We measure the performance attained by these implementations, on high-end hardware with a substantial number of cores.We explore the impact of the differences in algorithm, and also of the low-level implementation decisions. © Springer-Verlag 2013.
CITATION STYLE
Jung, H., Han, H., Fekete, A., Röhm, U., & Yeom, H. Y. (2013). Performance of serializable snapshot isolation on multicore servers. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7826 LNCS, pp. 416–430). https://doi.org/10.1007/978-3-642-37450-0_31
Mendeley helps you to discover research relevant for your work.