Deployment Topology View
A diagram showing where software actually runs — regions, zones, networks and hosts — which is where availability and cost properties become visible.
Logical diagrams show what the system is. The deployment view shows where it runs, and several critical properties are invisible until it is drawn.
What it makes visible: fault domains — whether redundant components are genuinely in different zones, or nominally redundant on the same rack; network boundaries and which traffic crosses them, which is both a security and a cost question; single points of failure that logical diagrams conceal, such as one NAT gateway, one shared cluster or one identity provider; and data residency, which for regulated workloads is a compliance statement.
It is the diagram that answers "what happens when this zone fails", and drawing it is frequently how a team discovers that the answer is worse than assumed.
What to include beyond the boxes: the routing path from the user to the workload, including DNS, CDN, load balancer and gateway; replication direction between data stores, which reveals RPO; and the control-plane dependencies required for recovery — because a design that depends on autoscaling to survive a zone loss depends on the control plane being healthy during exactly the event that may have degraded it.
Generate it from infrastructure-as-code where possible, since a hand-drawn deployment diagram is stale within weeks and is dangerous when it is.