Amazon S3 is not just a place to dump files. You can use it like a data warehouse as it is highly available and extremely fault tolerant. File durability is 99.999999999% over the span of a year. S3 is also a Hadoop-compatible filesystem. And you do not have to keep three copies of the data as it already has data replication built in.

S3 is not good for data that is updated frequently, but it is great for data that is appended frequently. IoT data tends to fall squarely in the latter category so S3 works well. Many companies that are advanced in analytics keep large datasets in S3 and create temporary Hadoop clusters for processing and analytics. Results are then stored, and the temporary Hadoop clusters are terminated to save costs.