We're making an MMO. We need a world, not a match.

Ravi has a whiteboard diagram of zone servers stitched by a home-rolled “world brain”. It works for 200 players in a single region. 5,000 players across two continents is the goal. The home-rolled option is a year of engineering he doesn't have.

PlayServ's engineer walks him through Realm: Iris renders zones, Realm owns the world model, Edge shards ownership across regions.

The existing Iris zone servers stay. Each zone becomes a Realm session; entities declared on Iris are bridged into Realm's entity model via a thin adapter. Ownership is explicit — the castle that straddles two zones has a single authority, not two.

A player running from Zone A to Zone B no longer sees a loading screen. Entity ownership migrates under 50 ms; the client receives a smooth stream of state from both zones during the handoff. Integration tests replay a recorded 2-hour raid across three zones. Zero desync events.

Edge adds a second region. Players pick the closest shard automatically; parties with mixed-region players are routed to a central pool. The studio's DevOps person — there's one — doesn't need to learn Kubernetes to bring up the second region.

The studio runs a synthetic 5,000-player stress test across five zones and two regions. Ownership transitions, zone handoffs, and cross-shard parties all hold. Ravi shelves the “home-rolled world brain” repo and redirects two engineers onto gameplay.