The ability to scale out, run in multiple data centers across different regions seamlessly is good enough reason alone! I've been using this product since its early days at Amazon Web Services where it was called AWS Sagemaker (also known as SageMaker). At that time we ran our machine learning models within containers which made them portable between compute instances without copying files or restarting other services like databases etc.. Now there's no need anymore - you can just spin up more of these container-based "computers" instead of having an EC2 instance running your model 24/7 when not needed by another application layer service such as RDS MySQL database server / Redis cache node etc... It takes some getting used too but once you get into flow things become much easier than they were before. We haven't had any issues so far aside from those related directly to cloud infrastructure itself e g., network congestion causing delays while spinning large numbers off machines etc.- nothing.