Posted On: Mar 20, 2023
We are excited to announce the support for Graviton3-based instances in Amazon EMR. You can now use Amazon EC2 C7G instances with EMR on EC2 and Amazon EKS. AWS Graviton3 processors are the latest in the AWS Graviton processor family. They provide better compute performance, floating point performance and support DDR5 memory that provides 50% more memory bandwidth as compared to DDR4. Amazon EMR launches support for Amazon Elastic Cloud Compute C7g (Graviton3) instances which improve cost-performance for Apache Spark workloads by up to 13%. You can improve cost-performance of Spark workloads running on EMR on EKS by up to 15% by migrating to Graviton3-based instances. The benchmark used in this post to arrive at the cost-performance improvement is derived from the industry-standard TPC-DS benchmark, and uses queries from the Spark SQL Performance Tests GitHub repo with the following fixes applied. Performance on your workloads may vary and therefore not yield similar cost savings.
In addition, EMR adds support for the following:
- Amazon EMR launches support for Amazon EC2 C6i, M6i, I4i, R6i and R6id instances which improve cost-performance for Apache Spark workloads by 6% to 33%
- Amazon EMR launches support for Amazon EC2 M6a, C6a instances which improve cost-performance for Apache Spark workloads by 15% to 50%
- Storage optimized Instances with up to 30 TB of NVMe SSD instance storage (lm4gn, ls4gen),
- Memory optimized instances with up to 1,952 GiB of DRAM-based instance memory (X1),
- Memory optimized and general purpose instances (r5n, M5, m5dn, r5ad, and m5ad).