Using Microsoft R in Enterprise Environments

  • tomaz.kastrun

    SSCrazy

    Points: 2085

    Comments posted to this topic are about the item Using Microsoft R in Enterprise Environments

    Tomaž Kaštrun | twitter: @tomaz_tsql | blog:  https://tomaztsql.wordpress.com/

  • David.Poole

    SSC Guru

    Points: 75199

    I enjoyed this article very much.

    The open source debate is an interesting one. OpenSSL is open source so any company that trades over the internet is trusting their most important interactions to open-source.

    Android is open source so it facilitates their confidential communications. Ios is based on BSD so there's no escaping it there either.

    For me open source means I can choose to look at the source code.

    It is neither a statement of quality or price. I've seen expensive open-source solutions and I've seen high quality open source solutions. The two categories don't necessarily collide.

    When talking about a model DB I think it is worth stressing that it is not the system model DB but a statistical model DB.

    As a DB guy with developer leanings I was initially confused by what data scientists and SAS jockeys refer to as a model. The main point of confusion was when I took delivery of a model and to me it looked like a collection of cross tabs. I had some interesting discussions about how the model would react to new data, how it would update, how long it would take to update etc. All amicable but a real eye opener

  • tomaz.kastrun

    SSCrazy

    Points: 2085

    Hi David,

    Thank you for pointing out the model DB is not a one of the system database, but a new database for modelling purposes.

    Well model can not only be a set of cross tabulations, as it does not embody any function / formula / method. I presume, they were relating to e.g.: confusion matrix - sort of cross tabs presentation of matrix holding the scores of model evaluation.

    Best, Tomaž

    David.Poole (8/26/2016)


    I enjoyed this article very much.

    The open source debate is an interesting one. OpenSSL is open source so any company that trades over the internet is trusting their most important interactions to open-source.

    Android is open source so it facilitates their confidential communications. Ios is based on BSD so there's no escaping it there either.

    For me open source means I can choose to look at the source code.

    It is neither a statement of quality or price. I've seen expensive open-source solutions and I've seen high quality open source solutions. The two categories don't necessarily collide.

    When talking about a model DB I think it is worth stressing that it is not the system model DB but a statistical model DB.

    As a DB guy with developer leanings I was initially confused by what data scientists and SAS jockeys refer to as a model. The main point of confusion was when I took delivery of a model and to me it looked like a collection of cross tabs. I had some interesting discussions about how the model would react to new data, how it would update, how long it would take to update etc. All amicable but a real eye opener

    Tomaž Kaštrun | twitter: @tomaz_tsql | blog:  https://tomaztsql.wordpress.com/

  • David.Poole

    SSC Guru

    Points: 75199

    Yes, when I first started working with SAS analysts I thought their output was going to be a bunch of formulae that I would have to code into an application.

    It came as a bit of a shock to be presented with the scores from a model.

    There was quite a bit of culture shock on both sides. DBAs learning the flexibility requirements of the data scientists and data scientists learning the requirements for 24/7 production systems. A good time was had by all as both sides relished the opportunity to learn and found each others disciplines fascinating

  • paolo.nadalutti

    Valued Member

    Points: 56

    As usual, very interesting posts Tomaž! 

    Recently I've been curious about in-database services update. 
    I see my in-database R version is stuck at 3.2.2, with a 8.0.0 Revolution packages version - the RTM SQL2016 version. As some R packages I use have evolved since R 3.2.2 and require R upgrade, I decided to test the in-database upgrading via SQL Server Service pack and Cumulative Updates. 

    After updating, I was surprised to find out that in-database R was not updated at all, and still at 3.2.2 - 8.0.0.

    So I decided to test the "binding" functionality, being aware that it comes only with Enterprise license. Anyway everything worked, and I was able to get a 3.3.3 R (maybe a newer one, I don't remember). The most surprising thing is that, when I "unbinded" my in-database R (as we usually don't work with Enterprise license), the whole in-database installation was removed!!! 

    So actually the question is: can I update in-database R, or should I adapt to an older version and modify my way to work?

    Thanks!!
    Paolo

Viewing 5 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply