[SOLVED] Is the difference between the system time of the Rumtime normal?

TeaYoung · 08-31-2015, 03:37 AM

Hello.

I have to questions regarding the System Clock. (# date)

It is it normal hours during the Runtime System is getting faster or slower?

I know it can be synchronized with the network time through the ntp or the rdate.

If you are not going to use the network system time is slow under normal circumstances it can not be avoided?

ericson007 · 08-31-2015, 04:14 AM

I am not sure exactlyhow time is kepton the kernel, but I am sure it must have something to do with the hardware clock. The hardware clock can skew over a period of time. Differences in time may be caused by your battery running flat.

Will be interesting to find out how all this is done.

TeaYoung · 08-31-2015, 04:36 AM

Quote:

Originally Posted by ericson007

I am not sure exactlyhow time is kepton the kernel, but I am sure it must have something to do with the hardware clock. The hardware clock can skew over a period of time. Differences in time may be caused by your battery running flat.

Will be interesting to find out how all this is done.

I know the details related to the hardware clock (RTC, RTC battery).

I know that you apply to the hardware clock with hwclock command to boot from the system clock.

Runtime is the time of system clock management in the Kernel would wonder whether the error is a natural phenomenon.

ericson007 · 08-31-2015, 05:07 AM

If my interpretation of what you are asking about is correct, then I believe it may be a factor of kernel tick frequency. So time scheduling within the kernel for interupts and tasks etc. I believe that indirectly this could skew the clock because probably that same system is what drives the time, hence the advice ntp servers must not be on virtual machines where the effect can be greatly magnified.

michaelk · 08-31-2015, 05:41 AM

Is this a real or virtual machine?

Not an expert but basically the kernel uses a timer to increment a counter which is the system clock. Depending on the hardware and other options like frequency scaling it can run slow or fast. There are ways to manually discipline your clock by using adjtimex but depending on the timer and how it drifts may not be practical.

https://the-b.org/Linux_timers

Basically the sole purpose of the hardware clock is to set the system clock at boot time.

TeaYoung · 08-31-2015, 08:24 AM

Quote:

Originally Posted by ericson007

if my interpretation of what you are asking about is correct, then i believe it may be a factor of kernel tick frequency. So time scheduling within the kernel for interupts and tasks etc. I believe that indirectly this could skew the clock because probably that same system is what drives the time, hence the advice ntp servers must not be on virtual machines where the effect can be greatly magnified.

Thank you for comment.
Would If I understand getting slower and System Time (date) Runtime on a Linux system during a long period of time? I am not a virtual machine

TeaYoung · 08-31-2015, 08:36 AM

Quote:

Originally Posted by michaelk

Is this a real or virtual machine?

Not an expert but basically the kernel uses a timer to increment a counter which is the system clock. Depending on the hardware and other options like frequency scaling it can run slow or fast. There are ways to manually discipline your clock by using adjtimex but depending on the timer and how it drifts may not be practical.

https://the-b.org/Linux_timers

Basically the sole purpose of the hardware clock is to set the system clock at boot time.

Thank you for comment.
I am not a virtual machine.
Like you and I seem to be able to slow or fast because of the frequency.
100% stable CPU frequency will not think it exists.
Thus, the stability of the System Time(#date) CPU frequency I think that this can slow down or speed up.
Do you also think so?

It has been helpful in your comments.

ericson007 · 08-31-2015, 04:35 PM

A while back i was reading up on getting the most accurate time for my idm. Ptp came up as really good but never bothered implementing it. Apparantly accurate to milliseconds. There was also talk of integrating gps hardware for accurate time. Thus on critical time systems, the drift experienced would be really milliseconds per day if not allowed to update. The hardware solutions though can get pretty expensive really quickly.

TeaYoung · 08-31-2015, 09:08 PM

Quote:

Originally Posted by ericson007

A while back i was reading up on getting the most accurate time for my idm. Ptp came up as really good but never bothered implementing it. Apparantly accurate to milliseconds. There was also talk of integrating gps hardware for accurate time. Thus on critical time systems, the drift experienced would be really milliseconds per day if not allowed to update. The hardware solutions though can get pretty expensive really quickly.

I can not even use the network even GPS.
System Clock is understood that vary depending on the hardward clock.
hardward clock goes faster system clock goes faster.
hardward clock is slow slow system clock.
My thought is that correct?

ericson007 · 09-01-2015, 12:34 AM

I think more correct would be the frequency of kernel access to the ticker for the clock, which in turn would likely be affected by cpu etc.