Cloudera Inc. today announced a preview of a new cloud-native machine learning platform that runs on Kubernetes, the popular orchestration platform for software containers.

Containers are portable, self-contained software environments that include code and all dependencies to able applications to run reliably in multiple computing environments.

The company said the new Cloudera Machine Learning platform will deliver fast provisioning and automatic scaling as well as containerized, distributed processing in heterogeneous computing environments. It’s intended to combine secure data access with a unified experience across on-premises, public cloud and hybrid environments. Secure data access spans Hadoop’s HDFS file system, cloud object stores and external databases.

The move represents a continuation of Cloudera’s move away from reliance on the Hadoop big data platform that gave birth to the company. As Hadoop has become a commodity service in the cloud, the company has been moving up the value chain by targeting machine learning as a core competency.

Cloudera said it’s seeing increased demand by enterprises to make machine learning part of their day-to-day operations. Cloudera Machine Learning is meant to enable those organizations to lower the barriers to machine learning development by enabling users to provision their own environments with minimal responsibility by information technology department staff.

The software expands upon the workflow elements of the on-premises Cloudera Data Science Workbench with cloudlike features such as automatic scaling, distributed dependency isolation and distributed graphics processing unit training. It can be installed into any supported Kubernetes environment using standard Kubernetes tooling, meaning there is no dependence on host processors. Dependency management is provided by containerized Python, R and Apache Spark-on-Kubernetes libraries.

The product is initially aimed at organizations that want to deploy machine learning primarily in the public cloud using public cloud storage services as well as customers who have an existing cloud-managed Kubernetes environment. The company said it plans to offer the product as a managed service in the future. Cloudera’s Data Science Workbench will continue to be the platform of choice for on-premises deployments.

Pricing wasn’t specified. The product is expected to ship next year. Companies can sign up for a preview here.

Image: Pixabay