OpI wrote two hands on articles about real work big data usage. The first is how to do data ingest with Apache Spark and Apache Zeppelin with code examples.
I also did research on Twitter's replacement for Apache Storm called Twitter Heron. Heron is compatible with Storm topologies and add some interesting tools and features.
Twitter has opened source another real-time, distributed, fault-tolerant stream processing engine called Heron. They see as the successor for Storm. It is backwards compatible with Storm's topology API.
There was a great webinar on SOLR6 updates including a deep dive into SQL and Graph support. SOLR 6 is really becoming a force. Running Apache Solr with Hadoop is a killer stack for searching massive datasets.
Kafka has a new release, 0.10.0 which includes KafkaStreams.
Google has released SyntaxNet, an open source neural network framework for TensorFlow that does Natural Language Understanding. It includes the awesomely named Parsey McParseface, an English parser that Google trained and can be used to analyze English text. For the name alone I have to investigate it and see if I can work it into a HDP / Spark machine learning pipeline.
SyntaxNet: Neural Models of Syntax — I am investigating it now, but this project requires some extensive builds and horsepower to prepare.
The Top Three Insights From the Apache Committers is fascinating and a good preview of next months Hadoop Summit in California.
The Apache Big Data Conference released a ton of great presentations and information updates:
Hadoop 3 will be add features to double effective storage capacity while increasing data resilency by 50 percent through code erasure. Hadoop 3 will be Java 8 only , add more than two name nodes, add native optimization. For smaller files, they will use 64kb stripes for better handling small IOT files. Here is a nice look ahead to Hadoop 3.
Apache Hive 2.0; SQL, Speed, Scale — Performance enhancements in Hive 2.0 with LLAP for fast queries, LLAP + Tez acceleration, Cost Based Optimizer improvements, now removes support for ancient Java 6, Hadoop 2.x only, Hive CLI is being deprecated. So big changes in Hive coming; updates all around and what should be rapid speed and capability improvements.
Streaming SQL with Apache Calcite — If you have not heard about Calcite it's been used by Apache Drill for a while. It's being used by everyone to improve their SQL. Apache Calcite includes a SQL parser, validator, JDBC driver, query optimization, relational algebra support and more. The list of projects powered by Calcite is impressive including Apache Flink, Apache Drill, Apache Hive, Apache Kylin, Apache Phoenix and Apache Storm. Calcite was previously called Optiq.
Protecting Enterprise Data in Apache Hadoop is extremely important as your data lake is the one source for your data. Hortonworks and Apache take this need seriously; so enterpise hadoop provides security via Kerberos, SSL, Apache Atlas, Apache Ranger and other projects.
Ozone - Object Store for Apache Hadoop Using Hadoop as an object store with REST API. With Ozone, you can replace S3 with your own in-house data store.
Under the Hood with Ambari Metrics and Grafana Scalable monitoring has been added and is not visualized with the very customizable dashboards in Grafana.
Parallelizing Genome Variant Analysis
This is a series of great new articles that show how to do genome variant analysis with Apache Spark and ADAM (a scalable API and CLI. This is about as complex science as you are going to find and married with complex Spark machine learning. But this shows you the power of big data and Spark processing. Again more amazing things coming out of AMPLab.
Spark 2.0 will ship with the second generation of Tungsten engine that includes volcano iterator model for improved query evaluation strategy.