Is a managed implementation of Apache Hadoop (but also Apache Spark, Apache HBase and Presto) that can run on EC2 (EKS, AWS Outposts - and since June 22 even serverless).

EMR abstracts and reduces the complexity of the infrastructure layer used with traditional MapReduce frameworks.

It is a big data platform designed for Extract, Transform, Load function ( ETLs) and for Machine Learning, Data Analysis and Business Intelligence.

EMR has some similarities with Glue in terms of ETLS ( but will outperform it for ML and Data Analysis, or be the only choice if workloads requires any other engine other than Spark)