Re: Small Redo Log File Size

When there is a lot of redo activity ('redo size' in v$sysstat) the files will fill up quickly and when the file is full one needs to switch to a new file. At that time all the dirty buffers will be checkpointed. So if you have a large buffer cache, small redo log files and a lot of activity then potentially many buffers will be written to disk. This will increase the amount of buffers written to disk (you write more buffers for checkpoints than for normal writes). This will help to limit the I/O bandwidth for the processes doing reads, so thre reads will potentially slow down.

Small redo logs were the default for the oracle database for the longest time, it is also one of the biggest things people don't change and it will impact performance for most databases.

Anjo.

Dereck L. Dietz wrote:
> I'm working with an Oracle 10g Version 1.0 datawarehouse database upgraded
> from an Oracle 9i version.
>
> While running the Enterprice Manager and browsing through the performance
> recommendations I came across one which stated the Redo Log Files were too
> small. The size of the current Redo Log Files (3 groups of only 1 file
> each) are only 10MB. Oracle recommended the size of 2GB for the Redo Log
> Files.
>
> Since we've experienced very slow response times the more activity there is
> against the database I think this should be taken seriously. I'm expecting
> resistance - or apathy - when I suggest following this recommendation.
>
> Since I'm not the "official" DBA there (and a contractor to boot) I'd like
> to have as much ammunition as possible to back up my suggestion.
>
> Would it be possible for the more knowledgeable on this group to post about
> the benefits of properly sized Redo Log Files and the pitfalls of improperly
> sized Redo Log Files? I'd like to be able to print them out to use if I
> have resistance to this recommendation.
>
> Thanks.
Received on Wed Aug 16 2006 - 04:03:51 CDT