Skip to content

Backpressure

Propagate downstream capacity pressure upstream so producers slow before overload accumulates into failure.

The Diagnostic Story

Symptom: One part of a coupled system keeps producing or admitting work faster than the part receiving it can absorb. The backlog grows invisibly, latency climbs, and the downstream side becomes less capable precisely when it needs headroom to recover. The upstream actors keep going because nothing tells them to stop.

Pivot: Make downstream capacity legible upstream. Propagate the stress or saturation signal against the direction of ordinary flow so that producers slow, pause, defer, or prioritize before accumulation becomes overload rather than after.

Resolution: Queue growth stays bounded, the downstream receiver holds enough slack to stabilize, and critical flow continues under prioritization rules. The system sacrifices some throughput and accepts higher latency, but preserves viability and the ability to recover without manual rebuilding.

Reach for this when you hear…

[stream processing] “The consumer can't keep up and we're just letting the queue pile up until the whole pipeline falls over.”

[manufacturing / kanban] “Assembly keeps pushing parts to the next station but the station is already full — we need to stop sending until they clear the line.”

[incident response] “We're getting paged faster than we can triage, and new alerts keep coming in while we're still buried in the first wave.”

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (6)

  • Constraint: Limits possibilities to guide outcomes.
  • Feedback: Outputs influence inputs.
  • Observability: Infer internal state externally.
  • Queueing: Organizes tasks into a waiting line based on arrival and service rates.
  • Signaling: Revealing hidden information.
  • Threshold: Safe vs harmful levels.

Also references 4 related abstractions