I use Zeppelin in different form and shape and it is very promising. Some colleagues are mentioning that Jupiter can do all that Zeppelin handles.
I have not used Jupiter myself. I have used Tableau but that is pretty limited to SQL.
Anyone has used Jupiter and can share their experience of it vis-à-vis Zeppelin?
Dr Mich Talebzadeh
Disclaimer: Use it at your own risk. Any and all responsibility for any loss, damage or destruction of data or any other property which may arise from relying on this email's technical content is explicitly disclaimed. The author will in no case be liable for any monetary damages arising from such loss, damage or destruction.
You might want to take a look at this:
I use both Zeppelin and Jupyter myself, and I would say by and large the conclusions of that article are still mostly correct. Jupyter is definitely superior in terms of stability, language (kernel) support, ease of installation and maintenance (thanks to conda) and performance. If you just want something that works well straight out of the box, then Jupyter should be your goto notebook solution. I would say this is especially true if your workflow is largely in python since many of the Jupyter developers also have close ties with the general python data analytics / scientific computing community, which results in better integration with some important packages (like matplotlib and bokeh, for example). This makes sense given that the project was originally a part of ipython after all.
However I definitely think Zeppelin still has an important place. The vast majority of Zeppelin users also use spark (also an apache project), and for that use case it should always be better than Jupyter given that its backend code is written in Java (a JVM language). There are also several advanced features that Zeppelin has that are somewhat unique, including a simple API for sharing variables across interpreters (https://zeppelin.apache.org/docs/0.7.0-SNAPSHOT/interpreter/spark.html#object-exchange). There's also the angular display system API (https://zeppelin.apache.org/docs/0.7.0-SNAPSHOT/displaysystem/back-end-angular.html). Granted, these two features are currently only fully supported by the spark interpreter group but work is currently underway to make the API extensible to other interpreters. Lastly, I think the most powerful feature of Zeppelin is the overall concept of the interpreter (in contrast to Jupyter's kernels) and the ability to use them together in a single notebook. This is my main reason for using Zeppelin since I regularly work with both spark/scala and python together.
So tl;dr, if you are using spark and/or have workflows which use multiple languages (namely scala/R/python/SQL), you should stick with Zeppelin. Otherwise, I would suggest Jupyter.
On Mon, Nov 28, 2016 at 5:06 AM, Mich Talebzadeh <[hidden email]> wrote:
Data Scientist I
Science Data Modeling and Computing (398K)
Jet Propulsion Laboratory
California Institute of Technology
Tel: <a href="tel:%2B1-818-354-6012" value="+18183937720" style="font-size:12px;color:rgb(17,85,204)" target="_blank">+1-818-354-6012
"Granted, these two features are currently only fully supported by the spark interpreter group but work is currently underway to make the API extensible to other interpreters"
On Mon, Nov 28, 2016 at 9:23 PM, Goodman, Alexander (398K) <[hidden email]> wrote:
Thank you guys for valuable inputs.
I have never used Jupyter myself but have used Zeppelin. Obviously it sounds like if the Big Data deployed has Spark centric view of things (with Spark being the penicillin of Big Data World :) together with Scala and SQL, then Zeppelin is a goof fit. I have also noticed recently that Hortonworks are actively promoting Zeppelin. However, I do appreciate that there are fans of Python around.
May be a strategy would to offer both. Having said that there are hard core users that would never give up on Tableau!
Dr Mich Talebzadeh
On 28 November 2016 at 20:32, DuyHai Doan <[hidden email]> wrote:
I can comment the reasons I use Zeppelin, though I haven't used Jupyter extensively. This is for a Fortune 500 company shared by many users.
-Easy to write new Interpreter for organization specific requirements (e.g. authentication, query limits etc).
-Already using Java and AngularJS extensively so it was a great fit.
-LDAP and Notebook level permissions worked great.
-Default D3.js visualization system works pretty well (could use some improvement)
-Easy to create and share business user friendly reports.
-Wide variety of Interpreters (JDBC, Spark, R, Mongo, custom etc).
-So far has been stable.
On Mon, Nov 28, 2016 at 12:59 PM, Mich Talebzadeh <[hidden email]> wrote:
Would you mind sharing details on this?
How Active Directory/LDAP and Kerberos are integrated with Zeppelin?
Dr Mich Talebzadeh
On 30 November 2016 at 11:26, Mich Talebzadeh <[hidden email]> wrote:
This page has examples for both Active Directory and LDAP:
On Kerberos it could be done for example through
export SPARK_SUBMIT_OPTIONS="--principal xxx --keytab yyy"
in zeppelin-env.sh as an example - that's how we do that.
Or as explained here -
Hope this helps.
On Wed, Nov 30, 2016 at 3:51 PM, Mich Talebzadeh <[hidden email]> wrote:
|Free forum by Nabble||Edit this page|