Also refered to today as Tiering. One needs to find out the " working data set". It is a percentage of a given work load where the reads and writes are higher as compared to older data. A rule of thumb is about 20% for a transcational Data Base.
By using "Tiering" your TCO for a storage solution whether primary or secondary can have a remarkable effect. Remember, not all your applications require sub-millisecond response at all times.
Example - If I arrive at a "WOrking Set" for all my work loads to be 1.5 TB, I will require only 2TB of useable drives with very high performance such as a NVMe, and provide for more economical drives such as SAS or NL-SAS ( high capacity) for my data that is outside of the " working data set".
Most storage solutions whether it is a 3-tier or a HCI architecture provides for "Tiering".
Note - HSM is an old terminology, however is still used by some vendors.