RE: performance impact of archivelog

From: Flado <>
Date: Tue, 18 Nov 2008 15:05:26 +0100
Message-ID: <>

Here's a discussion on archivelog vs. noarchivelog mode:

It is quite long and somewhat noisy - but I just read most of it so let me summarize:

A big (negative) performance impact of switching to archivelog mode is likely to come from the sudden importance of the NOLOGGING attribute. If you:
a) do lots of direct-path inserts,
b) your segments are LOGGING
Then switching to archivelog mode will cause a huge increase in redo generation: basically, your direct-path inserts will now be fully logged (where they previously weren't).

So if you do not have this setup, and you have taken measures to prevent the obvious problems (waits for log buffer space, log file switch completion (archiving needed), and so on) you should not notice the transition.


> I always see these things discussed in terms of "more" but never in terms
> of "how much".


What do you expect to see? The performance impact of archivelog varies, depending on too many things, from 0 (in most cases - see above) to infinity (when your archive log destination fills up). That's why the only true statement one can make about it - aside from "it depends(tm)" - is "possibly more".



Received on Tue Nov 18 2008 - 08:05:26 CST

Original text of this message