Re: partitionning - Mailing list pgsql-general

On Wed, 2005-03-09 at 13:21, Joshua D. Drake wrote:
> Scott Marlowe wrote:
>
> >On Wed, 2005-03-09 at 10:31, FERREIRA William (COFRAMI) wrote:
> >
> >
> >>hi
> >>
> >>does the table partitionning exists under PostgreSQL ? or maybe an
> >>alternative exists ?
> >>
> >>Oracle implements this system : it allows to spread rows on differents
> >>partitions depending on an attribute.
> >>For example, my application store xml documents in database, an the
> >>partitionning is used to spread the differents documents on differents
> >>partitions.
> >>
> >>
> >
> >Currently, doing this in PostgreSQL is more of a "roll your own" thing.
> >you create the base tables on different table spaces, then create an
> >updatable view with the proper triggers to make sure that the incoming
> >records go to the right tables.
> >
> >With the advent of very large raid arrays with very fast caching
> >controllers, this methodology is becoming less and less necessary.
> >
> >
> I don't agree with this. There are many postgresql users
> who have 100, 200, 600 GB databases that don't have the budget
> to purchase a 20,000 array.

A Good AMI Megaraid card with 512 meg of battery backed cache will work
the same here.  I.e. eliminate the need for partitioning.  it doesn't
have to cost $20,000 to do it.  And let's never forget that the data is
where the value is, not the hardware.   Notice I wasn't saying there was
never a need anyway, just that it's becoming less of an issue each day.
which it is. Back in the day a decent one scsi interface RAID card with
battery backed cache was >$2,000.  nowadays, they've dropped to the $300
to $500 range.  I can't see the cost of a DBA figuring out and
implementing partitioning schemes costing less than that in time.  Plus
it puts your data on a more reliable system.  Penny wise and pound
foolish to go cheap in my opinion.

Anyone keeping 600 gigs of data and trying to save $500 in hardware
costs is saving costs in the wrong places, imnsho...

> Table partitioning is a way to keep things efficient. That
> should be regardless of technology.

Yes, it is a way.  It's just a less necessary one than it once was, with
hardware now able to provide the same performance increase with little
or no work on the users part.  We've got to weigh the increased
complexity it would take to implement it in Postgresql and maintain it
versus the gain, and I say the gain is smaller every day.

> RAM is cheap, so lets just use as much of it as we can even
> if it means we swap.

What does table partitioning have to do with RAM?