Backpressure
Signaling upstream to slow down when a system can’t keep up, instead of collapsing.
Backpressure is how a system pushes load back when it’s saturated — bounded queues, rejecting requests, or slowing producers — so it degrades gracefully rather than running out of memory. It pairs with rate limiting and retries with backoff.
Worked example: a producer emits 10k msgs/sec but the consumer handles 2k/sec — without backpressure the queue grows unbounded until memory dies. Backpressure signals the producer to slow or blocks it, keeping the system within capacity. Gotcha: drop vs block vs buffer is a real choice — blocking stalls upstream, unbounded buffering just moves the failure; bounded queues plus a shed policy are the honest design.