Re: Inheritance efficiency

2010/5/1 Greg Smith <greg(at)2ndquadrant(dot)com>:
> Vincenzo Romano wrote:
>>
>> While I can agree that "Enterprise grade" is a buzzword, it does mean
>> something: "very large amount of data" among other.
>>
>
> http://thedailywtf.com/Articles/Bitten_by_the_Enterprise_Bug.aspx
>
> It's quite straighforward to get PostgreSQL up and running with many
> terabytes of data, so long as you respect the design trade-offs in some
> options.  What you can't do is say those are wrong and reject alternative
> implementation suggestions just because they're not "enterprise".  Whenever
> anyone uses that word at me, I mentally replace it with "super duper", and
>
>> There's no "fundamentally good design", but only a design which takes
>> limitations and constraints into account.
>>
>
> You mean like taking into account the fact that partitioning performance has
> an unavoidable trade-off, where you have to balance the query optimizer
> overhead of supporting many partitions against the improvement from
> splitting data into smaller pieces?

Or taking into account that some parts of the engine are not scalable.
Almost all current RDBMS can cope with terabytes.
Almost none (if any) can cope with large number of partial indexes (provided
they support them) or child tables or table level constraints and so on.
This is a fact as far as I've read so far.

Then we can discuss about the enterprise grade, the fault tolerance
and whatever else
buzzword pops up in our minds.
These ones maybe not.

--
Vincenzo Romano
NotOrAnd Information Technologies
NON QVIETIS MARIBVS NAVTA PERITVS