Skippy: a New Indexing Method for Long-Lived Snapshots in the Storage Manager

Download: pdf,

"Skippy: a New Indexing Method for Long-Lived Snapshots in the Storage Manager" by Ross Shaull(Brandeis), Liuba Shrira (Brandeis), and Hao Xu (Brandeis). ACM SIGMOD Conference, Vancouver, CA, June 2008.)

Abstract

The storage manager of a general-purpose database system can retain consistent disk page level snapshots and run application programs ``back-in-time'' against long-lived past states, virtualized to look like the current state. This opens the possibility that functions, such as real-time trend analysis and audit, formerly available in specialized temporal databases, can become available to general applications in general-purpose databases.
Up to now, in-place updating database systems had no satisfactory way to run programs over long-lived, disk page level, copy-on-write snapshots in real time, because there was no efficient indexing method for such snapshots. We describe Skippy, a new indexing approach that solves this problem. Using Skippy, database application code can run against an arbitrarily old snapshot, and iterate over snapshot ranges, as efficiently as it can access recent snapshots, for all update workloads. Performance evaluation of Skippy, based on theoretical analysis and experimental measurements, indicates that the new approach is effective and efficient, providing close-to-optimal access to snapshots at low cost.

BibTeX entry:

@miscellanious{skippy-paper-sigmod08,
   author = {Ross Shaull and Liuba Shrira and Hao Xu},
   title = {"Skippy: a New Indexing Method for Long-Lived Snapshots in the Storage Manager"),
   booktitle = {{\em ACM SIGMOD Conference 2008}},
   address = {Vancouver, Canada},
   month = June,
   year = {2008}
}