Jim C. Nasby wrote:
>On Thu, Jul 14, 2005 at 11:29:23PM +0200, Martijn van Oosterhout wrote:
>>On Thu, Jul 14, 2005 at 11:30:36AM -0500, Jim C. Nasby wrote:
>>>On Wed, Jul 13, 2005 at 07:52:04PM -0400, Bruce Momjian wrote:
>>>
>>>>This is a good point. We have always stored data on disk that exactly
>>>>matches its layout in memory. We could change that, but no one has
>>>>shown it would be a win.
>>>
>>[...]
>
> My thought was to convert as pages were read and written. That should
> minimize the code impact.
If that were practical, even more radical I/O saving tricks might be
possible beyond removing alignment bytes - like some compression algorithm.
Jim C. Nasby wrote:
> Or maybe as an alternative, would it be possible to determine how much
> space in a given relation was being wasted due to padding? That could be
> used to figure out how much IO could be saved on different transactions.
Well, I do notice that if I gzip my larger tables's data files they
tend to compress between 80-95% so it seems there's a fair amount of
redundancy in at least some tables.
Has anyone tried running postgresql on a filesystem that does compression?
It seems that would be an easy way to guess at the ultimate
potential I/O savings of separating memory layout from disk layout.
Ron