I imagine it could deliver a nice geo-distributed tx-log with fairly cutting edge low-latency (compared with any other cloud service I can think of), but I suspect the current xtdb-jdbc module design would require reworking a bit to really align with Spanner's internal model and get the full benefit. So if you really value geo-distributed low-latency writes then it might be worth exploring, but more generally speaking I can't imagine it's going to be cost-effective choice until Spanner's pricing drops dramatically.

Pricing isn't an issue for our case. I know a new adapter would need to be written for spanner, but what else might need to change?

Nothing else that I can think of. It really would just be a case of extending/testing/modifying/forking xtdb-jdbc 🙂

Spanner definitely doesn't like incrementing row keys. This is their recommendation for generating sequences: https://cloud.google.com/solutions/sequence-generation-in-cloud-spanner I'll take a look at see if this seems like it might work with XTDB. There's going to be a latency hit, but for our application this is acceptable.

Interesting - I hadn't looked at that page. The asynchronous sequence generator probably could work (not batch though)