Redshift is based upon PostgreSQL and is used for OLAP, online analytical processing. Data must be loaded into Redshift before you can run queries on data, however it is extremely efficient at data analysis once data is loaded, data can be store directly in Redshift but I would recommend storing data in RDS or S3. It utilizes columnar storage as opposed to row, which can store more data with the same amount of disk space and allows for more efficient queries. It has a pay as you go model as more and more instances deployed automatically. Similar to Athena you can use SQL for queries.
You can create a Redshift cluster. The primary node is in charge of planning out queries and the remaining nodes perform the actual queries. For DR purposes, Redshift clusters can be multi-AZ or you can utilize point in time snapshots, these are incremental so only changes are saved, saving on space. You can even have Redshift push these snapshots to another region if you need to perform disaster recovery in another region. Another interesting use case of the Redshift cluster once it’s created is Redshift spectrum. Let’s say you want to analyzed data in an S3 bucket but don’t wish to import the data into Redshift, by utilizing Redshift Spectrum your Redshift cluster with automatically launch spectrum nodes that will analyze the data directly on S3 and send the results back to the redshift cluster for final analysis and then the requester, saving time on analysis by not having to move manually.