Re: One datafile?

"Apparently on a standalone set up this has yielded performace benefits. The logic being that one big datafile means less work for I/O than 2. I personally doubt that would be much of a difference."

This statement is absolute hogwash. The best way to address I/O contention is to relocate resource-competing tables, indexes, datafiles, etc., to different drives, controllers, etc. And, with most major operating systems (including NT) supporting Async I/O, more drives means better performance. Period.

Tim Palmer wrote:

> A collegue of mine has suggested that we could test running our
> production database (V7.2 about to go to V7.3 on NT3.5 about to go to 4)
> by just using one big data file (system and user data all in together).
> We have a RAID 5 set up reading one logical drive - so contention and
> I/O concurrency is irrellevant.
>
> Apparently on a standalone set up this has yielded performace benefits.
> The logic being that one big datafile means less work for I/O than 2. I
> personally doubt that would be much of a difference.
>
> Does anyone have any thoughts on this. There is no requirement to take
> tablespaces offline individually for backups or whatever - we run in
> archivelog mode with a cold backup every night.
>
> The growth of the database is slow (and is in terms of data not objects)
>
> My first reaction was to this was that it would be a bad thing to do -
> but why not? Are there *really* any dangers associated with this
>
> On the other hand is there much to gain from this performance wise? I
> suspect not ??
>
> Tim Palmer
Received on Sat Aug 01 1998 - 08:54:25 CDT