Thank you. At this moment, I am considering this one: https://dataverse.harvard.edu/. It is recommended by the publisher I want to submit my paper to and it is clear who supports it (Harvard and others). On the other hand, there is very little on the topic my paper is about.
I think it's worth supporting if it's recommended by the publisher. I used figshare https://doi.org/10.6084/m9.figshare.5579092.v1 or Mendeley Data http://dx.doi.org/10.17632/jw5m448svx.1 and I did not notice any fundamental difference... importantly, datasets get unique identifiers.