Backpressure
A flow-control mechanism that limits upstream production when downstream capacity cannot keep pace, preventing uncontrolled queue and memory growth.
Backpressure constrains upstream work when a downstream consumer cannot process incoming load at a sustainable rate.
Why Queues Grow
If long-term arrival rate λ exceeds service capacity μ, the queue grows without bound. An unbounded in-memory queue can hide saturation temporarily while latency and memory use continue to increase.
Implementation Forms
Backpressure can be implemented by:
- blocking the producer,
- rejecting new work when a bounded queue is full,
- demand signaling from the consumer,
- shedding selected low-priority work under overload.
Different from Rate Limiting
Rate limiting often enforces an admission policy. Backpressure reacts to downstream capacity. The mechanisms can complement each other.
Unbounded retry can defeat backpressure by generating more work against an already saturated dependency.
Related: Queueing Theory, Tail Latency, Load Shedding, Safe Retry Design.