1 |
On Friday 07 March 2008, John J. Foster wrote: |
2 |
|
3 |
> Switching to 250Hz looks like it has solved the problem. No time lost |
4 |
> for a little over an hour now, and ntp is syncing properly, I think. |
5 |
> But my reading of the help on this setting led me to believe that |
6 |
> 1000Hz was right for a desktop system. |
7 |
|
8 |
More like "we think that 1000Hz *should* work better than 250Hz, but we |
9 |
don't really know for sure and YMMV..." |
10 |
|
11 |
> Can any explain what this setting actually does, and why it works |
12 |
> now? |
13 |
|
14 |
From /usr/src/linux/kernel/Kconfig.hz: |
15 |
|
16 |
Allows the configuration of the timer frequency. It is |
17 |
customary |
18 |
to have the timer interrupt run at 1000 Hz but 100 Hz may be |
19 |
more |
20 |
beneficial for servers and NUMA systems that do not need to |
21 |
have |
22 |
a fast response for user interaction and that may experience |
23 |
bus |
24 |
contention and cacheline bounces as a result of timer |
25 |
interrupts. |
26 |
Note that the timer interrupt occurs on each processor in an |
27 |
SMP |
28 |
environment leading to NR_CPUS * HZ number of timer interrupts |
29 |
per second. |
30 |
|
31 |
The timer wakes up x times per second and demands that it get attention. |
32 |
Your VM however, cannot control this on the host and the guest kernel |
33 |
interacts in strange ways with the host kernel. |
34 |
|
35 |
Just for interest, what are the Hz settings on host and guest? |
36 |
-- |
37 |
Alan McKinnon |
38 |
alan dot mckinnon at gmail dot com |
39 |
|
40 |
-- |
41 |
gentoo-user@l.g.o mailing list |