# Backpressure
Status: Growing
Backpressure is the ability of a system to slow down upstream work when downstream capacity is limited.
## Mental Model
If a downstream service, queue, database, or worker pool is overloaded, the upstream caller needs a signal to stop sending unlimited work.
Without backpressure, load keeps flowing into the system until queues grow, latency rises, memory fills, retries multiply, and the failure spreads.
## Why It Matters
Backpressure turns overload from an uncontrolled collapse into an explicit flow-control problem.
Examples:
- A queue stops accepting new jobs.
- A service returns `429` or `503`.
- A TCP receiver advertises a smaller window.
- A worker pool rejects work when all workers are saturated.
## Related
- [[Topics/Production Engineering]]
- [[Topics/Distributed Systems]]
- [[Field Notes/Tail Latency]]