The Apache Software Foundation has announced the latest version of Apache Ignite. Ignite is an open-source distributed database, caching and processing platform designed for transactional, analytical and streaming workloads. Apache Ignite 2.4 features new machine learning capabilities, Spark DataFrames, and the incorporation of a low-level binary client protocol.
According to the Ignite team, while a new version is normally rolled out every three months, Apache Ignite 2.4 was an exception and took five months in total.
The team says this release is a milestone release in that Machine Learning Grid is now production ready. ML Grid was created to enable users to run machine learning and deep learning training and inference directly on data stored within an Ignite cluster. In addition, it provides machine learning and deep learning algorithms specifically optimized for Ignite’s distributed processing.
“The component gained a variety of algorithms that can solve a myriad of regression and classification tasks, gave an ability to train models avoiding ETL from Ignite to other systems, paved a way to deep learning usage scenarios. All that now empowers Ignite users with the tools for dealing with fraud detection, predictive analytics, and for building recommendation systems…if you want. Note, ETL is optional, and the whole memory-centric cluster is at your service,” the team wrote.
ML Grid also recently received a software donation from NetMillennium that provided a form of genetic algorithms to solve optimization problems. “ The algorithms haven’t got to Ignite 2.4 and waiting for their time for a release in the master branch. Once you get them, you can apply the biological evolution simulation for real-world applications including automotive design, computer gaming, robotics, investments, traffic/shipment routing and more,” the team explained.
In addition, Spark DataFrames is now available and can be used to process data that is stored in an Ignite cluster. It can be constructed from a variety of sources, which include structured data files, tables in Hive, external databases, or existing RDDs.
Ignite 2.4 also comes with a low-level binary client protocol designed to expand the Ignite ecosystem. According to the team, this will provide communication with an existing Ignite cluster without starting an Ignite node. The application can connect through a raw TCP socket from any programming language.
“The beauty of the protocol is that you can develop a so-called Ignite thin client that is a lightweight client connected to the cluster and interacts with it using key-value, SQL, and other APIs,” said the team wrote
The full release notes are available here.