The Illusion of Simultaneous Events
In everyday life, time feels universal. If two people in the same room clap their hands simultaneously, everyone agrees it happened at the same moment. We share a single, continuous frame of reference.
In a distributed system, this intuition breaks down completely — and the consequences are severe.
Imagine a database cluster with nodes in New York and London. A user in Paris clicks a button that triggers a write to both nodes. Which write happened first? You might say, "Check the server timestamps." But this immediately runs into two fundamental physical problems:
- Clock drift: The quartz crystal oscillator inside a server is not perfectly accurate. It runs slightly fast or slow depending on temperature, age, and manufacturing variance. Two servers bought from the same vendor, racked side by side, will drift apart over time.
- Communication latency: Information cannot travel faster than the speed of light. A packet from Paris to New York takes at minimum ~35ms just from physics. Network congestion, routing, and queuing add unpredictable additional delay on top of that.
Because of these two facts, there is no concept of a single, global "now" in a distributed system. Two events on different machines cannot be perfectly ordered by physical time alone. This is not a solvable engineering problem — it is a fundamental constraint of physics. Everything else in this lesson is about how we work around it.