NTP Performance

[Achim Gratz]

Richard Laager via devel writes:
>> These may not yet have consistent TSC between cores/sockets (or require
>> BIOS tweaks for that).
> /proc/cpu says constant_tsc, but that's it (besides "tsc", of course).
> That is, I do _not_ have nonstop_tsc, so therefore I presume I do not
> have the "invariant TSC" CPU feature.

Constant TSC means that changing clock and certain PM states will not
alter the TSC frequency.   Nonstop TSC means that extends to "sleep"
states.  For many Intel processors, each core has it's own TSC and they all
start at slightly different times.  There is a special instruction to
read the TSC atomically together with some other per-core register that you can
set in order to determine which TSC you've been reading when the result
comes back.  Linux is supposed to set these extra registers up during
boot and change the readings so that all TSC give the same result.  If
you don't have a nonstop TSC that becomes more tricky as at least
individual sockets can and do sleep independently.  I am not sure which
families allow individual cores to sleep, but your server is using a
pretty old architecture.

> Any thoughts on what to look for in the BIOS? I poked around, but there
> didn't seem much related. There was a "Clock Spectrum Feature", which I
> assume is something about spread spectrum, which is disabled. The HPET
> is enabled. The Intel EIST setting is set to disabled, which the help
> text says disables C-states.

Not only that, it also should lock the frequency to a fixed one.

> Should I consider trying the HPET as the kernel clocksource?

I have no experience with that particular architecture, but I still
think HPET is at least an order of magnitude worse.  But it's only an
option on the copmmand line to change, so if you want to try…

> I only have the clock fuzzing errors on my NTP servers. I don't have an
> exact matching configuration that's not an NTP server, but: Similar
> hardware running Debian 10 and ntpsec 1.1.3 does not. Two eras of newer
> hardware running Ubuntu 18.04 and the same ntpsec do not.

I'm not really sure it is an error yet.  How many of these events do you
get?  Do things eventually stabilize?



Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

SD adaptation for Waldorf Blofeld V1.15B11:
http://Synth.Stromeko.net/Downloads.html#WaldorfSDada