Amazon EMR now supports Amazon EC2 G4 Instances which provides up to 4.5X faster and 5.4X cheaper XGBoost Training
Amazon EMR cluster running G4dn instances is 5.4X cheaper and 4.5X faster compared to an Amazon EMR cluster running Amazon EC2 R5 Memory-optimized instances. To learn more, please see Nvidia Blog Post.
You can now use Amazon EC2 G4 instances with Amazon EMR to take advantage of the latest generation NVIDIA T4 GPUs for machine learning and deep learning use cases. G4 instances are well suited for running machine learning inference applications such as image classification, object detection, recommendation engines, automated speech recognition, and language translation.
Amazon EMR enables you to run scalable machine learning workloads. Amazon EMR already supports Amazon EC2 G3 and P3 and instances optimized for running graphic intensive and compute intensive applications. You can use Amazon EMR to run deep learning and machine learning frameworks such as TensorFlow, Apache MXNet and add your preferred tools and libraries. Amazon EMR also supports EMR Notebooks and Apache Livy for building interactive Apache Spark applications.
For Amazon EMR pricing, please visit Amazon EMR pricing page. For documentation please review the Supported Instance Types, the Hadoop Daemon Configuration Settings, and Task Configuration.
Amazon EMR support for Amazon EC2 G4 instances is now Generally Available on EMR Versions 5.30 in the following regions: Asia Pacific (Hong Kong), Asia Pacific (Tokyo), Asia Pacific (Seoul), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), Canada (Central), Europe (Frankfurt), Europe (Stockholm), Europe (Ireland), Europe (London), Europe (Paris), Middle East (Bahrain), South America (São Paulo), US East (N. Virginia), US East (Ohio), AWS GovCloud (US), US West (N. California), and US West (Oregon). Here is a link to the overall Regional Availability of Amazon EMR.