I think aggregate-level deduplication is a massive saver and it compliments SSD, b'cos it means you will need lesser SSDs to store same amount of data. However, b'cos the dedupe is at the aggregate level, which means deciding whether to have dedupe enabled or not on a performance centric volume will no longer be a choice. Unless you have serious doubts about the future perfomance impact and stakes are high, then you will be forced to create a dedicated aggregate and put all the non-performance centric volumes on it.
I guess the game changer is the SSDs & Multi-core CPUs systems (AFF) for Aggregate-level-dedupe.
At this moment, the value of SSD for archiving is not there yet to move data to SSD.
In my circumstance, the volumes we are snapvaulting are expected to have over 7 year retention periods, and I think moving them to SSD does not provide sufficient value. I think it will be in the future, as SSD sizes are much larger than NL-SAS, so eventually, pricing per TB will decrease
Of course, we could move to other Ontap technologies, such as Altavault, but for now, I think snapvault provides a simple enough environment for us to think outside the box
Perhaps, b'cos of this issue NetApp came up with this tiering feature in 9.6, which enables you to keep hot data on a NetApp AFF system and to tier cold data to low-cost cloud tiers, which means a good ROI on SSDs for long term basis. But then the key word is 'cold' data.