Effects of Disabling Interrupts or Pre-Emption on Response Latency

Rate Monotonic Scheduling

Assumption

No resource sharing (processes do not share resources, e.g. a hardware resource, a queue, or any kind of semaphore blocking or non-blocking (busy-waits)).

Wikipedia “Rate Monotonic Scheduling”

Real world

We must protect shared resources with locks of some kind. The most aggressive:

  1. Disabling interrupts, and

  2. Disabling pre-emption.

What are the effects of real-time performance when this assumptions is violated?

Normal Interrupt Processing

../_images/normal_interrupt.png

Effect of Disabling Interrupts

../_images/disabling_interrupts.png

Effect of Disabling Pre-emption

../_images/disabling_preemption.png