Re: Parallel query execution

<br /><br /><div class="gmail_quote">On Wed, Jan 16, 2013 at 7:14 AM, Bruce Momjian <span dir="ltr"><<a
href="mailto:bruce@momjian.us"target="_blank">bruce@momjian.us</a>></span> wrote:<br /><blockquote
class="gmail_quote"style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> I mentioned last year that I
wantedto start working on parallelism:<br /><br />         <a
href="https://wiki.postgresql.org/wiki/Parallel_Query_Execution"
target="_blank">https://wiki.postgresql.org/wiki/Parallel_Query_Execution</a><br/><br /> Years ago I added
thread-safetyto libpq.  Recently I added two parallel<br /> execution paths to pg_upgrade.  The first parallel path
allowsexecution<br /> of external binaries pg_dump and psql (to restore).  The second parallel<br /> path does
copy/linkby calling fork/thread-safe C functions.  I was able<br /> to do each in 2-3 days.<br /><br /> I believe it is
timeto start adding parallel execution to the backend.<br /> We already have some parallelism in the backend:<br />
effective_io_concurrencyand helper processes.  I think it is time we<br /> start to consider additional options.<br
/><br/> Parallelism isn't going to help all queries, in fact it might be just a<br /> small subset, but it will be the
largerqueries.  The pg_upgrade<br /> parallelism only helps clusters with multiple databases or tablespaces,<br /> but
theimprovements are significant.<br /><br /> I have summarized my ideas by updating our Parallel Query Execution
wiki<br/> page:<br /><br />         <a href="https://wiki.postgresql.org/wiki/Parallel_Query_Execution"
target="_blank">https://wiki.postgresql.org/wiki/Parallel_Query_Execution</a><br/><br /> Please consider updating the
pageyourself or posting your ideas to this<br /> thread.  Thanks.<span class="HOEnZb"></span><br
/></blockquote></div>Honestlythat would be a great feature, and I would be happy helping working on it.<br />Taking
advantageof parallelism in a server with multiple core, especially for things like large sorting operations would be
great.<br/> Just thinking loudly, but wouldn't it be the role of the planner to determine if such or such query is
worthusing parallelism? The executor would then be in charge of actually firing the tasks in parallel that planner has
determinednecessary to do.<br /> -- <br />Michael Paquier<br /><a href="http://michael.otacoo.com"
target="_blank">http://michael.otacoo.com</a>