Welcome to the February 2016 edition of Log Compaction, a monthly digest of highlights in the Apache Kafka and stream processing community. Got a newsworthy item? Let us know.
We’ve been discussing many improvement proposals this month
KIP-41 – a proposal to limit the number of records returned by KafkaConsumer.poll method has been accepted.
KIP-42 – a proposal to add interceptors to producers and consumers has been accepted. This improvement creates interesting new monitoring options and once this is implemented, it will be interesting to hear how to community is using the new APIs.
KIP-43 and KIP-44 propose improvements and extensions to Kafka’s authentication protocols. These are still under active discussion, and if you are interested in security in Kafka, I suggest reading the wiki and the discussion to see where we are heading.
KIP-45 – a proposal to standardize the various collections that the KafkaConsumer API expects is still under discussion, with the benefits of more standardized approach being weighed against the desire to maintain backward compatibility for this new API.
Many of us are just learning the ins and outs of the new consumer. This recently published blog post, with a complete end-to-end example proves very useful.
A passionate developer wrote very detailed blog posts on Kafka integration with Spark Streaming. This includes the little-discussed question of how to write the results of the stream processing job back into Kafka.
LinkedIn wrote about new features in Samza. The blog post also includes sexy throughput numbers, description of their use-case and description of how Samza is used in their data products. Really cool stuff.
Google contributed their Dataflow API (but not implementation) to the Apache Software Foundation and are inviting other stream processing projects to implement their SDK. We are watching to see where this will take the active stream processing community.
Classic relational database management systems (RDBMS) distribute and organize data in a relatively static storage layer. When queries are requested, they compute on the stored data and then return results
Machine learning on real-time data is a powerful combination because you gain direct insights into your data, can make powerful decisions, and consequently improve your business processes and outcomes. It