Posted On: Nov 20, 2017
You can now use deep learning framework Apache MXNet (0.12.0), GPU instance types P3 and P2, and the AWS Glue Data Catalog with Presto on Amazon EMR release 5.10.0. Apache MXNet is a scalable deep learning framework that allows you to define, train, and deploy deep neural networks. Furthermore, you can preinstall custom machine learning and deep learning libraries on an Amazon Linux Amazon Machine Image (AMI), and create your Amazon EMR clusters with that AMI. Additionally, Amazon EMR now supports Amazon EC2 P3 and P2 instances, EC2 compute-optimized GPU instances, for deep learning and machine learning workloads. Also, you can now use the AWS Glue Data Catalog to store external table metadata for Presto instead of utilizing an on-cluster or self-managed Hive metastore.
You can create an Amazon EMR cluster with release 5.10.0 by choosing release label “emr-5.10.0” from the AWS Management Console, AWS CLI, or SDK. You can select MXNet to install it on your cluster, and you can launch your cluster with a custom Amazon Linux AMI by specifying a custom AMI ID. For pricing information about P3 and P2 instances on Amazon EMR, please visit the Amazon EMR pricing page. Please visit the Amazon EMR documentation for more information about release 5.10.0, MXNet 0.12.0, custom Amazon Linux AMIs, and using the AWS Glue Data Catalog with Presto.
Amazon EMR release 5.10.0 is available in all supported regions for Amazon EMR.