Martijn van Oosterhout wrote:
> Well, you get another issue, alignment. If you squeeze your string
> down, the next field, if it is an int or string, will get padded to a
> multiple of 4 negating most of the gains. Like in C structures, there
> is padding to optimise access.
Anecdotally I hear at least as many people say that their database
is more I/O bound than CPU bound; and it seems that adding bytes
for alignment is a way of reducing CPU for more disk I/O.
I guess unaligned access so expensive that it makes up for the extra i/o?
In response to
pgsql-general by date
|Next:||From: David Esposito||Date: 2005-07-13 13:50:46|
|Subject: Re: index bloat |
|Previous:||From: Gnanavel S||Date: 2005-07-13 13:42:54|
|Subject: Re: INSERT INTO from a SELECT query|