Apache Spark and Apache Zeppelin: What's Coming in HDP 2.4.2

In March 2016 we announced Apache Spark 1.6 GA on HDP 2.4 and provided the 2nd technical preview of Apache Zeppelin.

Since then, Apache Spark 1.6.1, a patch release with bug fixes, has been released by the open source community. Marching with the community, the upcoming maintenance release of HDP 2.4 will include Spark 1.6.1 GA and also provide the final technical preview of Apache Zeppelin.

The upcoming HDP 2.4.2 maintenance release will deliver the following features for Spark & Zeppelin:

• Certified SparkSQL with ODBC (ODBC driver available from Hortonworks).
• Bug fixes in Spark Oozie action for a Kerberos enabled cluster.
• Spark Streaming with Apache Kafka support in a Kerberos enabled cluster.
• SparkSQL & ORC performance improvements.
• Final technical preview of Apache Zeppelin that includes Kerberos support, LDAP Authentication, and identity propagation.

For Spark & Zeppelin security, performance and integration with Oozie and Kafka are quite relevant for enterprises and we have focused on delivering these features in this maintenance release.

At the 2016 Hadoop Summit in Dublin, one of our Zeppelin committers will outline the work in progress to bring multi-tenancy to Zeppelin.  Our Zeppelin, Spark, and security team members are also delivering a Spark & Zeppelin meetup in Dublin. This will be a good opportunity to get a peek into upcoming features and talk to developers.

What’s Next

The community is working on Spark 2.0, a major release that focuses on paying down architecture and design debt by:

• Removing obsolete packages.
• Other cleanup tasks like removing Bagel and Guava.
• Adding streaming line YARN options in Spark-Submit.

We expect Spark 2.0 to start the release vote later this month.

Apache Zeppelin is also making a lot of progress and recently R interpreters have been added. We continue to drive Zeppelin towards enterprise readiness and will deliver Zeppelin GA in HDP in our next HDP minor release. What’s next for Zeppelin includes:

• Run Zeppelin in a multi-tenant environment. We have opened a JIRA to track this work.
• Implement GIT support in Zeppelin for sharing Notebooks
• Enhance Zeppelin visualization and data import capabilities.
• Ambari managed Zeppelin LDAP Configuration

Recently we announced collaboration with Hewlett Packard Enterprise to accelerate the performance Apache Spark and we have made more progress on this plan. We will soon report back on progress on this front.

If you want to try out Apache Zeppelin on Spark, a tutorial is available here: http://hortonworks.com/hadoop-tutorial/apache-zeppelin-hdp-2-4/

Over the past year, we have had tremendous customer success with Apache Spark here at Hortonworks and we are starting to see the same interest materializing with Apache Zeppelin.  We continue to deepen our commitment and involvement within the Spark & Zeppelin communities and thank our customers, partners and Apache community in helping us in this journey.  Look out for Zeppelin technical blog in a couple of weeks, stay tuned.