 |
|
Full Paper (PDF)
|
Checkpointing is an important mechanism for limiting crash recovery times.
This paper describes a new checkpointing algorithm that was implemented inOracle 8.0.
This algorithm efficiently finds buffers which need to be written for checkpointing and easily scales to very large buffer cache sizes: it has been tested with buffer caches as large as six million buffers.
Based on this algorithm, we have implemented a new checkpointing mechanismwhich we refer to as the incremental checkpointing mechanism.
Incremental checkpoints are continuous, low overhead checkpoints that write buffers as a background activity.
Incremental checkpointing is able to continuously advance the database checkpoint, i.e., the starting position in the redo log for crash recovery, resulting in dramatic improvements in recovery time while imposing minimal overhead during normal processing.
The rate of buffer writes for incremental checkpointing can be controlled by the user to balance checkpoint writing overhead with recovery time requirements.
In this paper, we describe the new data structures and algorithms that have been implemented for checkpointing and for incremental checkpointing in Oracle 8.0.
|
@inproceedings{DBLP:conf/vldb/JoshiBLL98,
author = {Ashok Joshi and
William Bridge and
Juan Loaiza and
Tirthankar Lahiri},
editor = {Ashish Gupta and
Oded Shmueli and
Jennifer Widom},
title = {Checkpointing in Oracle},
booktitle = {VLDB'98, Proceedings of 24rd International Conference on Very
Large Data Bases, August 24-27, 1998, New York City, New York,
USA},
publisher = {Morgan Kaufmann},
year = {1998},
isbn = {1-55860-566-5},
pages = {665-668},
crossref = {DBLP:conf/vldb/98},
bibsource = {DBLP, http://dblp.uni-trier.de}
}
|
DBLP: Copyright ©1999 by Michael Ley (ley@uni-trier.de).
|
|
