We’ve implemented de-dup on primary storage using NetApp technology, works pretty well, but you have to be selective about the volumes you target. There seems to be a small overhead in performance, about 3% from what we’ve tested so far, but this has not been noticeable to end-users.
The CIFS volumes work pretty well if it’s unstructured data, I’ve not tried SAP, but suspect that since the data is structured you may not have quite the success, and since SAP may require higher performance I’d tend to avoid it
Using Netapp filers you must be sure that you are on a Data Ontap version that supports it. Only newer versions do and depending on the filer size the volume that can be deduped is limited in size. For example our 3040 filer can only dedup volumes smaller than 3TB.