Exciting New Features in 1.2, Flink-PPML, and Kafka Streams

https://a248.e.akamai.net/secure.meetupstatic.com/photos/event/9/1/9/c/600_452317276.jpeg

這次 Flink Meetup 分別邀請到了 IBM 與 Linker Networks 的講者分別分享 Streaming ML with Flink 與 Kafka Streams！另外，由於是第一次在台大資工舉辦，也會安排一場 Flink introductory talk 與介紹 Flink 1.2 最新的重點功能。

Agenda (3 sessions)

• Introduction to Apache Flink, and Exciting New Features in 1.2 - Gordon Tai, data Artisans

This will be an introductory talk about Apache Flink, its unique building blocks, and how it is driving advancements in the distributed data stream processing space. Some of the newest features in Flink 1.2 will also be introduced, including rescaling stateful streaming jobs, queryable state, and async I/O.

• Real-Time Streaming Prediction in Flink using Flink-PPML - Arey Liu, IBM

The predictive model markup language (PMML) is a widely used language to describe predictive and descriptive models as well as pre- and post-processing steps. That way it allows an easy way to export and import models from other ML tools. In this talk, I’ll introduce how to use R neural-network prediction model in Flink through Flink-PMML.

• Stream Programming with Kafka Streams - Jimmy Lu, Linker Networks

Stream programming has become a new programming paradigm to deal with a series of data or even infinite datasets. The presence of Kafka certainly speeds up the paradigm shift. In the context of big data, with the great help of stream processing frameworks like Flink, stream programming is far easier than before. However, in the context of microservices, even if we apply stream libraries like Akka Streams or RxJava for data manipulations, we still have to handle late events, make interactive queries and achieve exactly-once delivery and message-driven architecture by ourself until the introduction of Kafka Streams in the release 0.10.0. In this talk, I'll first explain what stream and stream programming are. And then introduce you Kafka Streams and how it facilitates stream processing in the context of microservices architecture. In the end, I'll point out the difference between Kafka Streams and Flink, and show you how they achieve exactly-once message delivery semantics.

Speaker Bios

Arey Liu

Arey is a Data Scientist at IBM Innovation Center.

https://a248.e.akamai.net/secure.meetupstatic.com/photos/event/5/d/e/a/600_458364042.jpeg

Jimmy Lu

Jimmy Lu is currently working at Linker Networks Inc. (http://www.linkernetworks.com/) as a data engineer. His primary focuses are building data infrastructure on top of SMACK stack and facilitating everything that troubles data scientists. He is recently getting more interest in machine learning technologies such as Spark ML and TensorFlow, trying to be hands-on at both data and engineering sides. Previously he worked for Digital River Inc., a global e-commerce solution provider and MicroMacro Mobile Inc., an e-ticket service platform as a software engineer.
@songyunlu (https://twitter.com/songyunlu)

https://a248.e.akamai.net/secure.meetupstatic.com/photos/event/5/6/f/6/600_458362262.jpeg

Tzu-Li (Gordon) Tai

Gordon is an Apache Flink Committer, and Software Engineer at data Artisans (http://data-artisans.com/).
@tzulitai (https://twitter.com/)

https://a248.e.akamai.net/secure.meetupstatic.com/photos/event/5/d/f/1/600_458364049.jpeg