Re: [Postgres-xc-general] Looks like co-ordinator is my perf bottleneck....somebody help...

[Nirmal S. <sha...@gm...>]

So if the coordinator time or total query time is the sum of time taken by query on each and every node then what is the use of having cluster. The whole point of having the cluster is to divide the work across different data node and reduce the query time by almost the no. Of data nodes.
I thought the overall time taken by query/coordinator is the max of time taken by all the nodes ( max( dn1,dn2....) instead of sum of time taken by all the nodes sum(dn1,dn2....)

Please let me know if my understanding is incorrect.

Nirmal

Sent from my iPad

> On Feb 1, 2014, at 10:19 AM, Sandeep Gupta <gup...@gm...> wrote:
> 
> Nirmal,
> 
>  Coordinator time is a function of sum of the output from each datanode. PGXC shows performance when the datanodes output small amount of data compared to original size.
> 
> -Sandeep
> 
> 
> 
>> On Sat, Feb 1, 2014 at 11:46 AM, Nirmal <sha...@gm...> wrote:
>> Hi Koichi,
>> 
>> My tables are not replicated. They all are distributed the way you explained.
>> For example, total record in one table is 600000 and i have 6 nodes so each and every node has got 100000 records.
>> 
>> Now the issue is that when I am running my query directly on data node it comes up in 5 sec and it is taking the same time on each and every node so it should take the same time if i run the query through coordinator but somehow instead on 5sec it's taking 22 sec. So somehow the query execution on nodes are happening correctly but data movement from nodes to coordinator is taking a lot of time.
>> 
>> Please advise.
>> 
>> Sent from my iPhone
>> 
>> > On Feb 1, 2014, at 12:17 AM, Koichi Suzuki <koi...@gm...> wrote:
>> >
>> > It is not a good way to replicate all the tables for write
>> > scalability.   The best way is to distribute transaction tables (very
>> > frequently written ones) and replicate master tables (less frequently
>> > written and frequently joined with transaction tables).
>> >
>> > Example is our DBT-1 benchmark.  Slide 12 of
>> > http://postgres-xc.sourceforge.net/misc-docs/20120614_PGXC_Tutorial_global.pdf
>> > shows how we designed DBT-1 table distribution for XC.  I hope this
>> > helps.
>> >
>> > Regards;
>> > ---
>> > Koichi Suzuki
>> >
>> >
>> > 2014-02-01 Nirmal Sharma <sha...@gm...>:
>> >> Hi,
>> >>
>> >> I created a pgxc cluster with one coordinator, one GTM and 6 data nodes (all
>> >> on same big machine ).
>> >> To test the performance, i ran one query through coordinator on my data
>> >> which is evenly distributed on all the nodes and it took total 25 sec to
>> >> complete.
>> >>
>> >> And then i ran the same query on datanodes directly and it took 5 sec on
>> >> each and every datanodes.
>> >>
>> >> Since query execution happens parallely on data nodes so ideally even if i
>> >> run the query through coordinator, it should not take more than 5-8 sec max
>> >> but i dont understand why is it taking 25 sec.
>> >>
>> >> Can somebody help me.?
>> >> Do i need to make some changes to my cluster configuration?
>> >>
>> >>
>> >> Regards
>> >> Nirmal
>> >>
>> >> ------------------------------------------------------------------------------
>> >> WatchGuard Dimension instantly turns raw network data into actionable
>> >> security intelligence. It gives you real-time visual feedback on key
>> >> security issues and trends.  Skip the complicated setup - simply import
>> >> a virtual appliance and go from zero to informed in seconds.
>> >> http://pubads.g.doubleclick.net/gampad/clk?id=123612991&iu=/4140/ostg.clktrk
>> >> _______________________________________________
>> >> Postgres-xc-general mailing list
>> >> Pos...@li...
>> >> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general
>> >>
>> 
>> ------------------------------------------------------------------------------
>> WatchGuard Dimension instantly turns raw network data into actionable
>> security intelligence. It gives you real-time visual feedback on key
>> security issues and trends.  Skip the complicated setup - simply import
>> a virtual appliance and go from zero to informed in seconds.
>> http://pubads.g.doubleclick.net/gampad/clk?id=123612991&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Postgres-xc-general mailing list
>> Pos...@li...
>> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general
>