Re: High load on interrupts
Hi,
On Sat, 27 May 2006, Ingo Juergensmann wrote:
> On Arrakis, one of my m68k buildds, I discovered recently that "top" shows a
> quite high activity doing hardware interrupts:
>
> top - 13:57:09 up 1 day, 21:52, 3 users, load average: 0.10, 0.07, 0.01
> Tasks: 69 total, 1 running, 68 sleeping, 0 stopped, 0 zombie
> Cpu(s): 0.6% us, 0.0% sy, 0.0% ni, 0.0% id, 0.0% wa, 99.4% hi, 0.0% si
> Mem: 127340k total, 116048k used, 11292k free, 28372k buffers
> Swap: 272192k total, 19532k used, 252660k free, 47608k cached
>
> This 99.4% hi seems to be something like an idle process counter as it drops
> relative to CPU usage:
This is what I get under 2.6.16:
top - 23:47:49 up 51 min, 1 user, load average: 0.01, 0.14, 0.26
Tasks: 32 total, 1 running, 31 sleeping, 0 stopped, 0 zombie
Cpu(s): 3.2% us, 1.9% sy, 0.0% ni, 94.3% id, 0.0% wa, 0.6% hi, 0.0% si
> Cpu(s): 2.5% us, 0.0% sy, 68.1% ni, 0.0% id, 0.0% wa, 29.4% hi, 0.0% si
> (vivaldi, another 060 m68k buildd)
>
> "vmstat 3" reports:
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 0 1 19528 10144 28468 48024 0 0 12 21 192 34 55 45 0 0
> 0 0 19528 10084 28468 48024 0 0 0 0 424 23 1 99 0 0
> 0 0 19528 10084 28472 48024 0 0 0 11 415 26 6 94 0 0
> 0 0 19528 10084 28472 48024 0 0 0 0 408 20 0 100 0 0
> 0 0 19528 10084 28476 48024 0 0 0 4 410 21 0 100 0 0
>
> So, roughly 400 interrups per timeframe.
Due to chained interrupts most interrupts are counted twice, so you get at
least the timer interrupt and vertical blank handler twice. It's maybe a
little weird but normal.
> Whereas on my PPC and i386 system:
>
> PPC:
> top - 14:04:13 up 10 days, 18:30, 5 users, load average: 0.00, 0.09, 0.08
> Cpu(s): 0.9% us, 0.3% sy, 0.0% ni, 98.3% id, 0.1% wa, 0.2% hi, 0.2% si
>
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 1 0 86412 36496 20 138348 0 0 0 0 29 301 0 0 99 0
> 0 0 86412 36496 20 138348 0 0 0 0 41 376 1 1 98 0
> 0 0 86412 36496 20 138348 0 0 0 0 33 328 1 2 97 0
> 0 0 86412 36496 20 138348 0 0 0 4 21 269 1 1 98 0
I think on ppc clock interrupts are not counted here (there are not
standard interrupts).
> i386:
> top - 14:11:32 up 49 days, 16:28, 9 users, load average: 0.03, 0.04, 0.00
> Cpu(s): 3.3% us, 1.5% sy, 0.0% ni, 94.7% id, 0.0% wa, 0.1% hi, 0.5% si
>
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r b swpd free buff cache si so bi bo in cs us sy id wa
> 0 0 19976 6952 28 228936 0 0 3 0 7 9 3 1 95 0
> 0 0 19976 6952 28 228936 0 0 0 0 287 401 4 1 95 0
> 0 0 19976 6952 28 228936 0 0 0 6 285 404 4 3 93 0
> 0 0 19976 6920 28 228940 0 0 0 9 350 431 5 5 90 0
> 1 0 19976 6920 28 228940 0 0 0 5 319 417 5 4 91 0
That would be a kernel with HZ=250.
bye, Roman
Reply to: