Comments

7 comments

  • Avatar
    Adrian Williams

    Hi Holger,

    The R node uses the R server functionality from the Rserve package to communicate with the target R environment. While Lavastorm has not tested the integration of Dataverse with the ValidR product, if ValidR is able to support an R server then it may be possible to run a R script from the R node within ValidR.

    The Mango Solutions website does not mention which packages are included in the ValidR product's pre-defined modules. They may be able to provide you with the details.

    Regards,

    Adrian

    0
    Comment actions Permalink
  • Avatar
    Holger Schimanski

    Can I also configure the target R environment for the standard nodes from Dataverse for K-means, Decision Forest, Linear Regression etc.?

    0
    Comment actions Permalink
  • Avatar
    Adrian Williams

    Hi Holger,

    It is not possible to configure the nodes in the predictive and statistical node pack to utilize an external R environment, they communicate directly with the embedded TIBCO Enterprise R Runtime that is deployed for use with Dataverse. This integration does not use Rserve.

    Best regards,

    Adrian

    0
    Comment actions Permalink
  • Avatar
    Holger Schimanski

    Can you think about adding flexibility here?

    I very much like the predictive and statistical node pack, but don't like that much the dependency to TIBCOs Enterprise R because it creates additional license costs and I maybe don't need the performance of TIBCOs R library and favor ValidR because of an existing environment. TIBCOs R library is providing same method signatures as open source R and therefor should be generally possible I guess.

    Best regards
    Holger

    0
    Comment actions Permalink
  • Avatar
    Adrian Williams

    Hi Holger,

    I will discuss specifics of this with our Engineering team. However, despite R/ValidR and TERR being very similar in the methods there are significant differences in the operation of the R node and the nodes in the predictive and statistical analytics node pack.

    The R node passes a monolithic R script to the R environment for processing as a single task where as the nodes in the predictive and statistical node pack leverage a Java API to interact with the TERR environment on a more granular basis - allowing the nodes to effectively submit individual R statements for processing and then react based on the response to each submission. There may be some R software license restrictions that mean we could not use such a tight linkage between our product software and an Open Source R/ValidR environment.

    Best regards,

    Adrian

    0
    Comment actions Permalink
  • Avatar
    Holger Schimanski

    Thank you for discussing this with your Engineering team! Does TERR has to run on the same instance as Dataverse for this tight integration?

    0
    Comment actions Permalink
  • Avatar
    Adrian Williams

    My current understanding is yes, it needs to be on the same machine - but I have asked for clarification.

     

    0
    Comment actions Permalink

Please sign in to leave a comment.



Powered by Zendesk