postgres-xc-committers Mailing List for Postgres-XC

Code-wise this is just reordering and rearrangement of code, so it should
not have any performance impact. But, yes this is going to pave the way for
fast query shipping, where we ship queries directly to datanode, when
coordinator is not required. The pgxc_planner() code as such was a bit
cluttered and was not readable. Hopefully it's better now.

Big blocker is the code in get_plan_nodes and examine_conditions_walker().
That's still something I am trying to understand.

On Thu, Sep 29, 2011 at 4:39 PM, Mason <ma...@us...>wrote:

> On Thu, Sep 29, 2011 at 6:41 AM, Ashutosh Bapat
> <ash...@us...> wrote:
> > Project "Postgres-XC".
> >
> > The branch, master has been updated
> >       via  dfa5b5171b09930bd6f2863bfb13cfd91c000d1f (commit)
> >      from  0dc793a19b61dba2122feb621ca640145cf2828e (commit)
> >
> >
> > - Log -----------------------------------------------------------------
> >
> http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=dfa5b5171b09930bd6f2863bfb13cfd91c000d1f
> >
> > commit dfa5b5171b09930bd6f2863bfb13cfd91c000d1f
> > Author: Ashutosh Bapat <ash...@en...>
> > Date:   Thu Sep 29 15:59:41 2011 +0530
> >
> >    Modularise pgxc_planner() function. This function first throws error
> for
> >    unsupported statements, then checks if the query can be completely
> shipped to
> >    datanodes (pgxc_fqs_planner()). If the later is not possible and
> coordinator is
> >    needed for evaluation of query, standard_planner() is invoked.
> >    Most of the code in pgxc_planner() is now moved to pgxc_fqs_planner().
> The later
> >    function doesn't plan a query if it needs coordinator for execution of
> query or
> >    the query does not need datanodes for evaluation. If the query can be
> completely
> >    evaluated on datanodes, pgxc_fqs_planner() creates the plan, which
> mostly
> >    contains only RemoteQuery node and ships most of the query to the
> datanode.
> >    EXECUTE DIRECT statements which are to be executed on datanodes, are
> also
> >    directly shipped to the datanode, without involving planner on the
> coordinator.
> >
> > M       src/backend/pgxc/plan/planner.c
> >
>
> This is more similar to how things were a few+ versions ago (back to
> the future).  It may give a small bump in performance. Have you done
> any tests to see if it helps performance for singleton statements?
>
> Thanks,
>
> Mason
>
>
> ------------------------------------------------------------------------------
> All the data continuously generated in your IT infrastructure contains a
> definitive record of customers, application performance, security
> threats, fraudulent activity and more. Splunk takes this data and makes
> sense of it. Business sense. IT sense. Common sense.
> http://p.sf.net/sfu/splunk-d2dcopy1
> _______________________________________________
> Postgres-xc-committers mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-committers
>



-- 
Best Wishes,
Ashutosh Bapat
EntepriseDB Corporation
The Enterprise Postgres Company

On Thu, Sep 29, 2011 at 6:41 AM, Ashutosh Bapat
<ash...@us...> wrote:
> Project "Postgres-XC".
>
> The branch, master has been updated
>       via  dfa5b5171b09930bd6f2863bfb13cfd91c000d1f (commit)
>      from  0dc793a19b61dba2122feb621ca640145cf2828e (commit)
>
>
> - Log -----------------------------------------------------------------
> http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=dfa5b5171b09930bd6f2863bfb13cfd91c000d1f
>
> commit dfa5b5171b09930bd6f2863bfb13cfd91c000d1f
> Author: Ashutosh Bapat <ash...@en...>
> Date:   Thu Sep 29 15:59:41 2011 +0530
>
>    Modularise pgxc_planner() function. This function first throws error for
>    unsupported statements, then checks if the query can be completely shipped to
>    datanodes (pgxc_fqs_planner()). If the later is not possible and coordinator is
>    needed for evaluation of query, standard_planner() is invoked.
>    Most of the code in pgxc_planner() is now moved to pgxc_fqs_planner(). The later
>    function doesn't plan a query if it needs coordinator for execution of query or
>    the query does not need datanodes for evaluation. If the query can be completely
>    evaluated on datanodes, pgxc_fqs_planner() creates the plan, which mostly
>    contains only RemoteQuery node and ships most of the query to the datanode.
>    EXECUTE DIRECT statements which are to be executed on datanodes, are also
>    directly shipped to the datanode, without involving planner on the coordinator.
>
> M       src/backend/pgxc/plan/planner.c
>

This is more similar to how things were a few+ versions ago (back to
the future).  It may give a small bump in performance. Have you done
any tests to see if it helps performance for singleton statements?

Thanks,

Mason

Project "Postgres-XC".

The branch, master has been updated
       via  dfa5b5171b09930bd6f2863bfb13cfd91c000d1f (commit)
      from  0dc793a19b61dba2122feb621ca640145cf2828e (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=dfa5b5171b09930bd6f2863bfb13cfd91c000d1f

commit dfa5b5171b09930bd6f2863bfb13cfd91c000d1f
Author: Ashutosh Bapat <ash...@en...>
Date:   Thu Sep 29 15:59:41 2011 +0530

    Modularise pgxc_planner() function. This function first throws error for
    unsupported statements, then checks if the query can be completely shipped to
    datanodes (pgxc_fqs_planner()). If the later is not possible and coordinator is
    needed for evaluation of query, standard_planner() is invoked.
    Most of the code in pgxc_planner() is now moved to pgxc_fqs_planner(). The later
    function doesn't plan a query if it needs coordinator for execution of query or
    the query does not need datanodes for evaluation. If the query can be completely
    evaluated on datanodes, pgxc_fqs_planner() creates the plan, which mostly
    contains only RemoteQuery node and ships most of the query to the datanode.
    EXECUTE DIRECT statements which are to be executed on datanodes, are also
    directly shipped to the datanode, without involving planner on the coordinator.

M	src/backend/pgxc/plan/planner.c

-----------------------------------------------------------------------

Summary of changes:
 src/backend/pgxc/plan/planner.c |  264 +++++++++++++++++++++++++--------------
 1 files changed, 169 insertions(+), 95 deletions(-)


hooks/post-receive
-- 
Postgres-XC