Robust Stream Processing with Apache Flink and Flink-HTM

These talks will happen on June 27th at the San Jose Convention Center the night before Hadoop Summit begins. All are welcome!

There will be a networking event for everyone attending the meetups starting at 5PM and then the talks will begin at 6PM. Drinks will be provided at the networking event.

Agenda

5:00-6:00PM Networking & Drinks (courtesy of Hortonworks)

6:00-7:00PM Talk: Robust Stream Processing with Apache Flink

7:00-8:00PM Talk: Flink-HTM

Where: Room LL21E

Abstracts:

Jamie Grier will talk give a talk and hands-on demonstration of some of the advanced features available in Apache Flink. The talk will focus on features unique to Flink that allow one to achieve truly robust computation over data streams.

Topics covered will be:

• Event Time vs Processing Time and why it matters
• Robust handling of state
• Handling failures
• Handling code or cluster upgrades without losing state
• Apache Flink's Savepoints
• Handling data re-processing after code changes, etc

Eron Wright will present flink-htm (https://github.com/nupic-community/flink-htm), a library for streaming anomaly detection and prediction using Apache Flink, based on Hierarchical Temporal Memory (HTM) (http://numenta.org/#theory) and implemented by Numenta's HTM.java (https://github.com/numenta/htm.java) library. We'll show some examples of using flink-htm, then dive into its internals to illustrate how to extend Flink with new operators and connectors.

Speaker Bios:

Jamie Grier is now Director of Applications Engineering at data Artisans where he’s extremely excited to be able to help others realize the potential of Flink in their own projects. His goal is to help others design systems to solve challenging problems in the real world.

Prior to joining data Artisans Jamie worked on the streaming computation team at Twitter and prior to that on streaming processing systems at Gnip.

Eron Wright is a Director of Engineering at EMC and an active HTM community member.