Achivelog Mode Performance

Question

Hi to all
My Production DB Specs
Oracle 11g Enterprise Edition (11.2.0.3)
Windows Server 2012 Standard (64 bit)
HP ProLiant DL360p Gen8
Processors Intel Xeon CPU E5-2620 0 @2GHz
16GB RAM
Total disk space 745GB
Database size 74GB
I would like to enable the Archivelog mode in my db. I understand the importance of having archive logs. However as the nature of every db is different, my concern is:
1] What are those configurations that I need to test first BEFORE implementing archivelog mode? (redo log settings, transactions used per day etc.?)
2] How can I reduce the impact on my running production database performance as much as possible, when I implement archivelog mode?
3] What are the things that I need to check and monitor AFTER implementing archivelog mode? (for example, Archivelog disk spaces?)
Thank you for sharing with me your precious experiences.
Edited by: moslee on Mar 27, 2013 11:53 PM

Answer 1

Before turning on archivelog you need to :-
- check what volume of log you write each day, and if there are peak periods (batches etc)
- check for log switch interval and consider if the log size is optimal
- define what your archivelog backup and delete policy will be (now long to keep on disk, how often to backup, how many backups to take before deleting)
- write and test your backup / delete scripts (rman ?)
- check that you have sufficient disk to handle unexpected events
- check that you have monitoring enabled to detect disk filling up
- consider using FRA instead of archiving to a directory
- consider whether to archive locally, or to a remote server or NFS share
- check for objects or transactions with nologging (select force_logging from v$database)
- consider if you will duplex archivelogs or not
- consider how many archiver processes you will need (log_archive_max_processes)
- consider what you want to happen if archival fails (log_archive_min_succeed_dest)
The ARCH processes only copy files, so impact on performance can easily be simulated, and will depend on the volume of files being copied, and number of parallel copies.
Archivelogs are useless unless you have a database backup, so you need to include that in your scripting and testing, along with restore tests.