Timescale has predictable cost anchored in familiar reality; it is driven by storage volume and system load. You don't have that familiar, awful cardinality problem that is common to _every_ tag-set-series data model system. It performs very well and predictably. It's really awesome. It's open source and self-hosting is easy: It is postgresql. You already know what self-hosting implies from that one statement and whether you're willing to do it. If you're not, you can pay Timescale to do it for you with Timescale Cloud. In my experience, Timescale Cloud was very effective for the months my team used it. Their community is great, and the Timescale maintainers actually address issues reported by the community (including me personally)! It was a welcome 180 degree change from the seemingly antagonistic stance certain other related open source projects take toward their community. Their people are really good.

The feature set (especially cont queries and the sql extensions)

When I first started evaluating time series databases, Timescale was already on my list. What I love about them is, 1. Natively built atop Postgresql, so one gets the best of both worlds 2. One can choose between their self-hosted, managed and cloud flavours 3. Excellent support and success teams that make sure you are set up, are good to go and help you with queries quickly 4. Excellent community, especially on slack, where you can ask/answer questions and support each other