Don’t forget to mirror it, to the redundancy level of the rest of VDEVs, and ideally you want to use different uncorrelated SSDs. You don’t want them wearing out at the same time.
1TB is too much, but will allow you to store more small files along with metadata on an SSD. This is configurable on dataset granularity.
Old thread, but I found something interesting about write amplification.
https://youtu.be/V7V3kmJDHTA
In short:
That zfsSlog result is super misleading. It’s not that less data was written overall… just SOME WAS WRITTEN TO ANOTHER DEVICE! It’s like if I told you I’ve reduced waste and showed you my half-full garbage can as proof (but not the trash I threw out the window 
).
When SSDs were small, amplification was something to care about. But today with much larger drives being common: you could 2x/3x-or more the writes and still not hit an average drives rated TBW in its useful lifetime.
(I know this is off-topic. And is more a general comment on SSDs than ZFS. SSDs are friggin’ amazing and a deal for the capabilites they give you. Don’t fear wearing them out.)
