A method and system for maintaining a desired service level for a processor receiving excessive interrupts. The method includes the operation of defining an interrupt processing period during which interrupts will be measured for a processor. The amounts of time spent by the processor during the interrupt processing period in interrupt context can be measured. Another operation is detecting an interrupt storm occurring for the processor based on the amounts of time spent by the processor in interrupt context. The interrupts received by the processor can then be restricted for a period of the processor's total processing time when an interrupt storm has been detected.