Défis des systèmes distribués

Splitting an app into independent services adds power but also complexity: distributed systems mean services talking over a network, which is never free.

Network latency is the time data takes to travel between services. It is always present, compounds when one request hits many services, and can only be minimized.

A partial failure is when some services fail while others keep running. Unlike a monolith's all-or-nothing crash, it can leave a transaction half-done.

With data spread across services and databases, data consistency gets hard: making every service see the latest value at once is far tougher than with one database.

Concurrency is many services touching the same data at once. Without care it causes race conditions, where timing decides the outcome and data can corrupt.

Observability is knowing what your system is doing right now. Tracing a request across services and gathering logs, metrics, and traces takes a real strategy.

Debugging distributed systems is exponentially harder: logs scatter across services, one request may touch 10+ of them, and errors are tough to reproduce.

Operational complexity grows fast: deploying dozens of services, scaling each by demand, and monitoring every component demands serious automation.

The Two Generals' Problem shows why network communication is fundamentally unreliable: over a lossy channel you can never be sure your last message arrived.

Let's test your understanding of the common challenges in distributed systems.

You mapped the key distributed system challenges: latency, partial failures, consistency, concurrency, observability, debugging, and operational complexity.