Amazon EMR is easy to setup and use. There are plenty of ways to get started, such as the AWS EMR console, or you can automate the whole thing using AWS command line tool awscli, or the boto3 api for Python. We use a combination of awscli and boto3 for automation. It provides best in class tools built in, and integration with other amazon services such as s3 for data as well as log aggregation, etc. It also provides a way to use AWS EC2 Spot instances, which reduce the cost to run Map Reduce jobs by 50-80% on average.
There are no major dislikes about EMR, but in general, they could provide more options to monitor the cluster, and also provide ways to troubleshoot the failed jobs and provide ways to recover failed jobs. But Amazon is going in the right direction and hope they address those things as well.
If you are on AWS and looking to run Hadoop/Spark, look no further than Amazon EMR, and it might save a lot of time in setting up your cluster, and helps you focus more on the application business logic than worrying about the infrastructure configuration and setup.
We use EMR primarily for data analytics and big data processing using Spark, Hadoop and we also use S3 for storing the output.