Federated queries get a bad reputation when a naive engine pulls whole tables across the network and joins them locally. SemanticFed is built around a cost-aware planner that does the opposite.
It pushes filters, column projections, and aggregations down to each source, so a warehouse does the heavy aggregation it is good at and only the small result crosses the wire. The planner estimates cost across sources and picks the path that moves the fewest bytes.
The result is federation that is fast and economical, not a tax — and a clear, inspectable query plan so engineers can see exactly what ran where.
An illustrative pre-launch scenario describing the planner being built.
This is a planner-architecture story, not a separate page you configure. The cost-aware behaviour is intrinsic to query execution, so there is no clickable "Do it yourself" web walkthrough for it.
Ready to make this your story?
