Producer-consumer model is very helpful to decouple the system components. However, the situation is quite often in which producer produces jobs more rapidly than consumers can consume them. It's a challenge to manage a large number of unconsumed jobs.
Back-pressure is one of the strategies that can mitigate the problem of quick-producing meeting slow-consuming.
Back-pressure is a feedback mechanism that allows systems to sense the burst job load and gracefully respond to caller.
Some common-seen strategies includes:
503 Service Unavailablewhen Web / RESTful Service is under heavy load.
The system need to degrade itself when under heavy load. Back-pressure can help way from such case.
Rx Observables are responsive objects listening on queue-alike data structures. They could be overwhelmed by other fast-producing Observables.
See more Backpressure throttling algorithm on how to fine-tune the parameters of Rx operators.
Applying back-pressure is one effective technique to handle high-load. We tend to handle already accepted requests and transactions first, and reject those can't be handled. If you have many components and the producing-consuming speed might mismatch, consider back-pressure.