This claims that Spanner commonly has to wait upto 100ms for an update. Tracking down specific numbers is harder than I expected but in general seem to be closer to 50ms and even single digits for some operations. Does anyone know more about Spanner performance?
Otherwise this seems like dank fire. I’m a bit reluctant to believe that AntidoteDB is “industrial grade” yet.
The Spanner paper itself has performance data:
It might have improved or worsened since then. Looked good to me given what it does. Only problem is that big delay with TrueTime on geographically-distributed ones. Gotta make sure the clients can tolerate it.