In my opinion there is no clear answer to this questions. All depends on your data and on the operations you perform. We run ~530 backup jobs in our K8s cluster on different types of web applications with storage between 100 GB and 2 TB and as you can see in the graphs there have all different characteristics in memory and CPU usage:
Cache directory behaves roughly the same as ram consumption, so we adjust this roughly in proportion. In our backup runs, we exec backup, check and restore commands. When you run a single backup command only, the resource consumption could be significantly lower.
In general, we don’t set CPU limits in our deployments, cause of unpredictable behavior in applications,