Thread: question on parallelism

question on parallelism

From
Andrew McIntyre
Date:
does postgres have this capability? specifically local intrapartition?
http://pic.dhe.ibm.com/infocenter/db2luw/v10r5/topic/com.ibm.db2.luw.admin.partition.doc/doc/c0004557.html

sorry for all the newbie pg questions...
--


Re: question on parallelism

From
Chris Travers
Date:



On Wed, Jan 1, 2014 at 7:35 PM, Andrew McIntyre <amcintyre@m-m.com> wrote:
does postgres have this capability? specifically local intrapartition?
http://pic.dhe.ibm.com/infocenter/db2luw/v10r5/topic/com.ibm.db2.luw.admin.partition.doc/doc/c0004557.html

Nope.  Here's a quick breakdown of what is done in parallel and what is done sequentially in PostgreSQL.

Separate queries run in parallel, and can piggy back on eachother's sequential scans.  In essence all of the work that is done in parallelism affects parallel queries.

There is no intra-query parallelism however.  This means that partitioning doesn't generally help here (although sequential scans on the same partition by different queries can run in parallel with good results, a single query cannot run several scans on different partitions in parallel).

If you need this sort of feature you are going to either need to go to a related product (like Postgres-XC) or you are going to need to have some other layer that can do the parallelism.

Best Wishes,
Chris Travers
 
sorry for all the newbie pg questions...



--
Best Wishes,
Chris Travers

Efficito:  Hosted Accounting and ERP.  Robust and Flexible.  No vendor lock-in.

Re: question on parallelism

From
Michael Paquier
Date:
On Thu, Jan 2, 2014 at 2:29 PM, Chris Travers <chris.travers@gmail.com> wrote:
>
>
>
> On Wed, Jan 1, 2014 at 7:35 PM, Andrew McIntyre <amcintyre@m-m.com> wrote:
>>
>> does postgres have this capability? specifically local intrapartition?
>>
>> http://pic.dhe.ibm.com/infocenter/db2luw/v10r5/topic/com.ibm.db2.luw.admin.partition.doc/doc/c0004557.html
>>
> Nope.  Here's a quick breakdown of what is done in parallel and what is done
> sequentially in PostgreSQL.
>
> Separate queries run in parallel, and can piggy back on eachother's
> sequential scans.  In essence all of the work that is done in parallelism
> affects parallel queries.
>
> There is no intra-query parallelism however.  This means that partitioning
> doesn't generally help here (although sequential scans on the same partition
> by different queries can run in parallel with good results, a single query
> cannot run several scans on different partitions in parallel).
>
> If you need this sort of feature you are going to either need to go to a
> related product (like Postgres-XC) or you are going to need to have some
> other layer that can do the parallelism.
In the case of XC, it is worth noting that a query can take advantage
of parallelism on a sequential scan only roughly for sharded tables.
In the case of replicated table, query is usually simply pushed to one
singke remote node, making the scan occur on this single node.
Regards,
--
Michael