On Thu, Aug 4, 2011 at 8:33 PM, Sumeet Jauhar <sumeet(dot)jauhar(at)gmail(dot)com> wrote:
>
> [ Sumeet ] ok so i agree we need to move ahead and shift to a higher
> version . But how do we decide that . Which one would you say is the
> stablest version of Postgres [ still supported version ] out in the market
> below beacuse Brad here says his 8.1 version did have performance impacts
> . Brad - How had you decide on the version . Was it the latest version
> available at that point in time or there was someother reason ? I am also
> pretty sure that upgrading 2 times would not have been easy .
I would upgrade to either 8.2 or 9.0 and here's my reasons. with 8.2
you still have implicit casts, which your application may depend upon.
Most other changes between 7.4 and 8.2 were pretty small, so if
you've got a lot of implicit casts in your SQL, 8.2 will be the least
painful of the upgrades to late model pgsqls. HOWEVER, 8.2 is getting
pretty old now and performance wise 9.0 will pretty handily beat it.
In terms of stability, there are no reports of any versions after
about 8.1 or 8.2 being particularly unstable, but keep in mind that
support for 8.1 and 8.2 will be ending / may have ended already, so if
you can possibly test against 9.0 and see if it works well enough,
then you should really do so. The changes to things like autovacuum
getting multi-threaded (8.3) HOT updates (8.3) on disk tracking of
free space map (8.4) and a few other big breakthroughs make going to
the latest (9.0) or near latest (8.4) much more attractive. And trust
me, you WILL feel the difference in performance, it's huge from 7.4 to
8.3 and after that the incremental changes are noticeable, if not as
huge.