Evaluating Toree

[Namit Kabra <na...@in.ibm.com>]

Hi,

Our team is creating a framework that would be a repository of Data
Analysis algorithms. These algorithms should eventually be executed on
Spark. So I am creating a REST API that would trigger these algorithms on
Spark and get the results.

There are several solutions that I am looking at, including...

      Apache Spark Hidden REST API
      Spark feature - Provide a stable application submission gateway in
      standalone cluster mode


While doing this I was exporing Toree. Do you think Toree will be a better
option? If yes, how do I get started with creating a simplistic REST call
to Spark using Toree.


Thanks and Regards,
- Namit Kabra
----------------------------------------------------
IBM India Software Labs,
http://namitkabra.wordpress.com/

Re: Evaluating Toree

[Chip Senkbeil <ch...@gmail.com>]

If you want REST functionality, using something like Cloudera's Livy might
be more useful: https://github.com/cloudera/livy

Apache Toree is geared more towards dynamic applications that have a need
to run a variety of jobs against Spark that are configured on the fly such
as custom filtering and transformations that a LoB user might do in a
dashboard to get more insight into his or her data. Toree works well with
Jupyter notebooks and you can use it as the backend to the
https://github.com/jupyter/kernel_gateway project, exposing websocket
access to Apache Toree.

If your algorithms are static, you are not collaborating with others to
tweak those algorithms, and you just want a RESTful API, Apache Toree might
not be the best fit.

We have demonstrated in the past using Apache Toree (before it was brougnt
into Apache) to power RESTful servers, interactive applications, and other
Spark-related tools:
https://www.youtube.com/watch?v=2AX6g0tK-us

On Thu, Dec 15, 2016 at 10:44 AM Namit Kabra <na...@in.ibm.com> wrote:

>
>
> Hi,
>
> Our team is creating a framework that would be a repository of Data
> Analysis algorithms. These algorithms should eventually be executed on
> Spark. So I am creating a REST API that would trigger these algorithms on
> Spark and get the results.
>
> There are several solutions that I am looking at, including...
>
>       Apache Spark Hidden REST API
>       Spark feature - Provide a stable application submission gateway in
>       standalone cluster mode
>
>
> While doing this I was exporing Toree. Do you think Toree will be a better
> option? If yes, how do I get started with creating a simplistic REST call
> to Spark using Toree.
>
>
> Thanks and Regards,
> - Namit Kabra
> ----------------------------------------------------
> IBM India Software Labs,
> http://namitkabra.wordpress.com/
>