Vaibhav Jain
← All notes
rough ·

Load Shedding

Dropping or rejecting work intentionally so a system can remain healthy under overload.

#systems#reliability

Load shedding means intentionally rejecting or dropping some work when the system is overloaded.

It sounds bad, but it can be better than accepting everything and failing for everyone. A system that sheds load can stay responsive for the requests it does accept.

Common examples:

  • returning 429 Too Many Requests,
  • dropping low-priority events,
  • timing out expensive requests,
  • rejecting new jobs when a queue is full.