Re: How to REINDEX in high volume environments?

On 28 Sep 2002 at 17:51, Justin Clift wrote:

> Shridhar Daithankar wrote:
> <snip>
> > Looks like we should have a subdirectory in database directory which stores
> > index.
>
> That was my first thought also, but an alternative/additional approach
> would be this (not sure if it's workable):
>
> - As each index already has a bunch of information stored stored for
> it, would it be possible to have an additional column added called
> 'idxpath' or something?
>
> - This would mean that the index location would be stable per index,
> and would allow for *really* high volume environments to keep different
> indexes on different drives.

I have to disagree.. Completely.. This is like turning PG-Metadata into
registry...

And what happens when index starts splitting when it grows beyond 1GB in size?

Putting indexes into a separate subdirectoy and mount/link that directory on a
device that is on a separate SCSI channel is what I can think of as last drop
of performance out of it..

Just a thought, as usual..

I don't know how much efforts it would take but if we have pg_xlog in separte
configurable dir. now, putting indexes as well and having per database pg_xlog
should be on the same line. The later aspect is also important IMO..

Bye
Shridhar

--
VMS, n.: The world's foremost multi-user adventure game.