Google's Spanner used atomic clocks and GPS to make a global database consistent.

A globally-distributed database wants transactions that look like they happened in one order, but servers in different datacenters do not share a clock and cannot agree on which write came first.

Spanner solves this with TrueTime: each datacenter reads from GPS receivers and atomic clocks, and reports a time interval instead of a single instant, so the uncertainty is explicit. A transaction waits until its commit time is safely after the latest possible time of any overlapping transaction, then commits with that timestamp to all replicas.

The result is the first system to offer externally-consistent distributed transactions at global scale, letting Spanner provide non-blocking reads in the past, lock-free read-only transactions and atomic schema changes across all replicas.

Make time-measurement trustworthy, not just fast

GPS and atomic clocks bound the uncertainty, so time can be trusted

Waiting out the uncertainty turns a guess into a provable ordering

One global timestamp gives all replicas the same commit point

It makes cross-region reads and schema changes safe to run in parallel

When a distributed system needs a total order, clever messaging isn't enough; make the shared clock trustworthy so ordering falls out of a time everyone agrees on.

Spanner became the backbone behind Google's cross-region transactional services and was later offered publicly as Cloud Spanner. Its TrueTime approach became a case study in using real-world clocks to solve distributed consistency.