Indexing alternatives for multiversion locking

Abstract

In this paper, we present several options for extending single-version indexing schemes for use with multiversion two-phase locking. The proposed options are orthogonal to the underlying indexing structure (e.g., hashing or B+ trees). The options differ in where they place references to individual versions; this version selection information is placed either with the data or in the leaf entries of one or more indices. We also briefly summarize the results of a performance study showing that storing version selection information with the data is usually the best option because it keeps indices smaller, thus enabling a larger fraction of the indices to remain cached in the buffer pool.