Are there any downsides of using Jupyter vs Zeppelin notebook (part of HDP stack) for Spark ?

[#1]

I have used Jupyter notebook before. Looking into if I am losing much goign with Jupyter instead of Zeppelin ?

[#2 · Accepted solution]

@Swapan Shridhar- Read below forums as to which one suits better in which situation:

https://www.quora.com/Should-I-use-Zeppelin-or-Jupyter-notebooks-for-data-analysis

https://www.linkedin.com/pulse/comprehensive-comparison-jupyter-vs-zeppelin-hoc-q-phan-mba-

[#3]

The second link no longer works. It would be nice to have a comprehensive comparison, rather than "jupyter is good for running python locally, zeppelin is better for cluster workloads"

[#4]

Updated the link. It was not picking up the - at the end.

[#5]

"Are there any downsides..?"

- in Jupyter you have to decide upfront what language each notebook is going to be in. But in Zeppelin you can switch between languages within a single notebook, passing variable values between the languages. This allows you immense flexibility. For example, you could use Scala to calculate some results in Spark and then display the results in JavaScript/HTML /Angular using your own custom visualizations. You can use the best tool for the job if you're comfortable working in more than one language.

Disclaimer: I've hardly used Jupyter and I know it has cell 'magics' which I think allow you to write individual cells in a language different to the one assigned for the whole notebook. However I don't know if you can pass data between cells of different languages in the same notebook. The issue at https://github.com/ipython/ipython/issues/4386 suggests not (although that is almost a year old now): "ipython/jupyter will not be supporting multiple kernels for a single notebook with variables being passed around between them, so I am closing this issue." For me, this is a big downside for using Jupyter.