liveBook · Manning

Backpressure

As explained in Fast by Default

Backpressure is a mechanism by which an overloaded component signals that upstream producers must reduce, reject, or delay work.

Backpressure lets an overloaded component signal that it cannot accept more work. With admission control, new requests are rejected or delayed rather than entering an unbounded queue that later times out. section 5.3.1

Fast rejection preserves acceptable latency for admitted requests. Trying to serve every request during overload instead degrades latency for everyone and can end in queue overflow. section 5.3.1

Concurrent-request limiting

Rate limiting can cap concurrent requests per user as well as requests per second. Slow connections then consume a concurrency allowance, while lower-priority traffic can be shed first to preserve critical payment operations. section 5.3.1

Make overload visible with operational measurements:

How Fast by Default covers it

How deep it goes
Explains overload rationale and a concurrent-request limiting example, without specifying a general implementation design.
Examples
Uses slow client connections, priority traffic shedding, and payment processing during incidents.
What you can do after
Operate/configure services by measuring queue depth and concurrent requests to expose overload.
How it connects
Connects backpressure to admission control, rate limiting, queue depth, concurrent requests, and cache stampede.

Also explained in

Compare how 10 books explain backpressure →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Fast by Default. The section links point to the parts of the book it draws on.