Talk:Interrupt storm

Hard Disk in PIO mode

Does anyone have expertise on "interrupt storm" issues being related to the hard disk being accessed in PIO mode? Such information could be a useful addition to the article.

I experienced an "interrupt storm" issue recently with my PC: it apparently was caused by Microsoft Windows' function of reducing IDE read-speed for the hard disk when CRC (disk read) errors are detected. This reduction is done step-wise with each error problem, as the system tries to find an error-free level of speed: first to lower Ultra-Direct Memory Access (U-DMA) modes, if available, and then finally to Programmed Input-Output (PIO) mode, which requires a lot more CPU time than the DMA method and can cause noticeable system slow-downs and lags (presumably the symptoms of "interrupt storms"). I think this is frequently the "interrupt storm" issue for users, judging from the number of forums discussing it (as revealed via Google search).

The Microsoft KB article IDE ATA and ATAPI disks use PIO mode after multiple time-out or CRC errors occur explains the fix for the issue of DMA degrading to PIO. Many of the aforementioned forum discussions reference that KB article.

If I have it right, then someone with the technical expertise to describe the issue accurately may want to add this info. to the main Interrupt Storm article, if the community deems it appropriate. Memetics (talk) 12:24, 15 August 2009 (UTC) [Edited for clarity and better talk-page contextualization: Memetics (talk) 13:29, 26 August 2009 (UTC) ]

Novice Questions

Interrupt storms seem like bad juju. I have a number of questions that the article doesn't really speak to.

• How likely is it to get interrupt storms from non-maskable interrupts? Can this kind of situation lead to a crash? What's a normal CPU utilization percentage for interrupts? WHat is the typical variation? What's a typical number of interrupts per second?
• How significant would a series of mouse move events be towards an interrupt storm? Are mouse events typically rate-limited? What would be a typical mouse rate?
• Are there other adverse impacts to interrupt storms than perceived slowness? What happens when interrupts are missed? How is servicing of interrupts prioritized?
• Does an RTOS manage interrupts in a way to prevent storms? If so, how?
• When does polling become more efficient than interrupts?

Are any of these questions of relevance to the article? 70.247.162.60 (talk) 05:29, 8 November 2015 (UTC)