Pgsql: Create an infrastructure for parallel computation in PostgreSQL

postgresql.org

165 points·amitlan··15 comments

Lots more info in the README

http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitd...

Exciting!

Back in 2013, I wrote this blog post [1], it's getting closer, but I should probably change the title to "Will PostgreSQL 9.6 Bring Back Moore’s Law?"

[1] http://joelonsql.com/2013/04/20/will-postgresql-9-5-bring-ba...

This is more of an Amdahl's Law thing though.

Can someone explain to me (and others) a potential usecase for this ?

Intra-query parallelism for starters.

that clears it up. thx

Will this be compatible with say pg_shard?

Does anyone know how this stacks up against CitusDB?

Is the target for Postgres to natively do what CitusDB does right now?

This is about parallelism on a single machine.

Thanks for clearing that up.

Are there any examples how to use this feature?

Note that this in itself is not a feature but part of the infrastructure for developing parallel execution nodes (for an example, you could take a look at parallel seq scan under development[1])

EDIT: [1] is a huge discussion, by the way.

[1] http://www.postgresql.org/message-id/CAA4eK1KTv73uD9_W5wR4Hi...

Thanks for the reply. Then, my question becomes as follows: Is there a way for basic PostgreSQL users (as me) benefit from this (yet)? I don't want to dive into any codes :)

Yeah, I think this is all under the hood, so in future releases, Postgres should perform even better for you ;)

It's a framework for future functionality. End-users won't be touching it for some time I expect. The best thing about psql is their careful, iterative engineering.

We might see parallel seqscan soon.

Reply on news.ycombinator.com