Fallacies of Distributed Systems #2: Latency is Zero

- Common pitfalls in building scalable systems and how to avoid them
- The basics of messaging and how it makes your software more reliable and maintainable
- How to architect long-running processes (hint: no more batch jobs!)
- How to sell the concept of messaging to your organization

===============

Udi Dahan explains how latency affects distributed systems. When we run code on our own machine, we assume (and it is often the case) that latency is effectively zero. But when we build a more geographically distributed system, overall system slowness increases by orders of magnitude at each step.

As an illustration, if 1 CPU cycle is scaled to be 1 second (in order to make the differences in magnitude easier for our brains to grasp), crossing the continental United States from San Francisco to New York takes 4 years!

That's what's so dangerous about our programming models. Declaring a dependency on some service can hide massive amounts of latency without the developer being explicitly aware of it.

===============