Dear all,

I investiguated a little more. I guess that the partitionBy primitive impacts a lot my performance since my data are not well balanced (70% of my tuples are assigned to a single task). It looks like the cluster mode takes many time to root the data.

Denis

Le 28/05/2015 21:28, Jeffery Maass a écrit :
Denis:

I would avail myself of Nathan Leung's offer to help. His statement was that a factor of 20 is much too great a difference. He asked for more details.

Thank you for your time!

+++++++++++++++++++++
Jeff Maass <mailto:maas...@gmail.com>
linkedin.com/in/jeffmaass <http://linkedin.com/in/jeffmaass>
stackoverflow.com/users/373418/maassql <http://stackoverflow.com/users/373418/maassql>
+++++++++++++++++++++


On Thu, May 28, 2015 at 10:51 AM, Denis DEBARBIEUX <ddebarbi...@norsys.fr <mailto:ddebarbi...@norsys.fr>> wrote:

    Thanks for your help.

    Denis


    Le 28/05/2015 15:15, Jeffery Maass a écrit :
    Did you know that a horse can race a human and the human will win
    -- in the first several yards?  Spread out the race longer and
    the horse will always win.

    The tests being run aren't covering enough situations.  Storm was
    designed to take business code and scale it out in order to
    handle massive amounts of data.  Try increasing your records per
    minute by a thousand fold and see how local mode functions versus
    cluster mode and business code spread out over multiple nodes.

    Thank you for your time!

    +++++++++++++++++++++
    Jeff Maass <mailto:maas...@gmail.com>
    linkedin.com/in/jeffmaass <http://linkedin.com/in/jeffmaass>
    stackoverflow.com/users/373418/maassql
    <http://stackoverflow.com/users/373418/maassql>
    +++++++++++++++++++++


    On Thu, May 28, 2015 at 2:46 AM, Denis DEBARBIEUX
    <ddebarbi...@norsys.fr <mailto:ddebarbi...@norsys.fr>> wrote:

        Dear all,

        I measured the performance of my topology written in pure
        storm (no trident). I compared it in local mode and then in
        cluster. The two experiements were done on the same single
        server (deployed with docker in cluster configuration). The
        result was that *the local mode is 20 times faster than in
        cluster configuration.*

        I am surprised about it. What could I do to improve the
        performance on the cluster?

        Thanks for your help.

        D. Debarbieux


        ------------------------------------------------------------------------
        <http://www.avast.com/>   

        L'absence de virus dans ce courrier électronique a été
        vérifiée par le logiciel antivirus Avast.
        www.avast.com <http://www.avast.com/>






    ------------------------------------------------------------------------
    <http://www.avast.com/>       

    L'absence de virus dans ce courrier électronique a été vérifiée
    par le logiciel antivirus Avast.
    www.avast.com <http://www.avast.com/>






---
L'absence de virus dans ce courrier électronique a été vérifiée par le logiciel 
antivirus Avast.
http://www.avast.com

Reply via email to