Evaluation of Google Maps' Design

Evaluate how the proposed mapping service behaves under load, stale traffic data, and segment-server failure. Partitioning helps distribute work, but the query must still return a valid route across partition boundaries.

Requirements compliance

Check the architecture against the stated workload and 2-to-3-second route budget. The earlier multi-million-RPS calculation was a stress scenario, not a demonstrated result.

Availability

Putting all graph data and request handling on one server creates several risks:

• Memory limits: The working set may exceed available RAM. Disk-backed processing remains possible, but its latency must be measured.

• Connection limits: A single server could not maintain persistent connections for millions of concurrent users.

• Single point of failure: If the server failed, the entire system became unavailable.

Segments let us distribute graph data. Availability also requires replicas and a way to route around failed hosts: