Re: PostgreSQL Timeline

On Thu, Oct 10, 2013 at 4:39 AM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com>wrote:

> On Thu, Oct 10, 2013 at 2:48 PM, Chris Travers <chris(dot)travers(at)gmail(dot)com>
> wrote:
> > Postgres-XC probably did more than anything else to kill a lot of the
> > clustering projects. Why use GridSQL when you can use Postgres-XC?
> Bookmarking this one :)
>

Because they are intended for different workloads. GridSQL handles ad hoc
analytical queries much better.

Try running DBT-1 (TPC-H) against GridSQL and Postgres-XC. You will get
nice scalability with GridSQL, but may be waiting for hours with
Postgres-XC... it sometimes ships everything to one single node for joining.

Postgres-XC does fine for OLTP workloads and analytical queries that do not
involve inter-node joins. GridSQL performs poorly for OLTP, but can handle
inter-node joins in parallel for analytical queries.

Use the right solution depending on your requirements.

That said, long term Postgres-XC could replace GridSQL if it improves in
this area. Postgres-XC could also replace pgpool-II today in some cases,
possibly more if we allow reads from data node slaves.

Regards,

Mason