Ensure that all Amazon Elastic MapReduce (EMR) clusters provisioned within your AWS account are using the latest generation of instances in order to get better performance at lower cost.
This rule can help you work with the AWS Well-Architected Framework
This rule resolution is part of the Cloud Conformity Cost Management tool for AWS
By using the latest generation of Amazon Elastic MapReduce instances instead of the previous generation of instances you can upgrade your EMR clusters for better hardware performance (faster CPUs, additional RAM memory, superior I/O and higher network throughput) at lower costs. For example, the new generation memory-optimized (R3) instances are 9% faster than the previous ones and the compute-optimized (C3 and C4) instances are 37% faster than the old generation (C1) instances. On top of all these performance improvements the latest generation instances are cheaper than the old ones, e.g. a c3.xlarge instance provisioned by AWS EMR in the US East region costs $0.263/hour whereas an old generation c1.xlarge instance costs $0.640/hour.
The following table (named EMR Previous Generation Instance Types) will help you to identify any previous generation EMR instance types in use:
|Instance Family||Previous Generation Instance Types|
|General Purpose||m1.small | m1.medium | m1.large | m1.xlarge|
|Memory Optimized||m2.xlarge | m2.2xlarge | m2.4xlarge | cr1.8xlarge|
|Compute Optimized||c1.medium | c1.xlarge | cc2.8xlarge|
|Storage Optimized||hi1.4xlarge | hs1.8xlarge|
The following table will help you choose the equivalent current generation instance type required for the AWS EMR hardware upgrade process:
|EMR Previous Generation Instance Types||EMR Current Generation Instance Types|
|m1.small | m1.medium | m1.large | m1.xlarge||t2.small | m3.medium | m3.large | m3.xlarge|
|c1.medium | c1.xlarge | cc2.8xlarge||c3.large | c3.xlarge | c3.2xlarge|
|m2.xlarge | m2.2xlarge | m2.4xlarge | cr1.8xlarge||r3.large | r3.xlarge | r3.2xlarge | r3.8xlarge|
|hi1.4xlarge | hs1.8xlarge||i2.4xlarge | d2.4xlarge|