Why PC clock isn't accurate?

[Phil]

I would install a lightweight linux in a virtualbox VM (e.g. debian + lxde or lubuntu), enable "bridged networking" in Virtualbox and try to get a time server working.
There's a package simply called "ntp", and "ntp-doc" with some help in it, well I guess it's the kind of help from Unix manual

Errr, what? I think when he said the PC wasn't connected to the internet, a virtualized Linux wouldn't be hooked up to the internet either.

But yeah, I have a localized NTP daemon running on my linux gateway which all my local clients synch to and the gateway itself that syncs with an online server periodically. I think smartphones typically sync by default to the network provider, if not explicitly deactivated (I think most phones need this to be activated for most synch operations to work, like synching to Gmail services etc).

 

[Albuquerque]

I would install a lightweight linux in a virtualbox VM (e.g. debian + lxde or lubuntu), enable "bridged networking" in Virtualbox and try to get a time server working.

While a good idea in theory, do not virtualize a time server unless you want everything under it to be permanently wrong. By its very nature, virtualization will cause irregular time drift. http://support.ntp.org/bin/view/Support/KnownOsIssues#Section_9.2.2.

Always install a time SERVER deamon on physical hardware. The client daemon is a good idea in virtual environments, although some hypervisors are better at this (Xen) than others (VMware and Hyper-V.)

 

[Blazkowicz]

While a good idea in theory, do not virtualize a time server unless you want everything under it to be permanently wrong. By its very nature, virtualization will cause irregular time drift. http://support.ntp.org/bin/view/Support/KnownOsIssues#Section_9.2.2.

Always install a time SERVER deamon on physical hardware. The client daemon is a good idea in virtual environments, although some hypervisors are better at this (Xen) than others (VMware and Hyper-V.)

I love that kind of nitpicking and caveat.
You can try the software side at least.

Errr, what? I think when he said the PC wasn't connected to the internet, a virtualized Linux wouldn't be hooked up to the internet either.

But yeah, I have a localized NTP daemon running on my linux gateway which all my local clients synch to and the gateway itself that syncs with an online server periodically. I think smartphones typically sync by default to the network provider, if not explicitly deactivated (I think most phones need this to be activated for most synch operations to work, like synching to Gmail services etc).

No need for internet, you can get the time from your watch, a smartphone, watching a shadow at solar 12 AM
The time server (cheap and low power computer) would keep that time (adding its own drift) and deal it to the LAN's PCs, who get about the same time. It does not seem too bad.

But ultimately if it's that important then the best while keeping cheap enough is to use a small computer with a GPS receiver attached as a time server. You get perfectly good time from the GPS, job is done.

 

[function]

Why is that happening? Can't they use a more accurate crystal or something? Why my cheap Casio is a lot more accurate even vs an expensive gadgets?

Laaaaaaate to the thread.

A watch that's normally on your wrist can perhaps be calibrated to keep time accurately at around body temperature. If the temperature remains constant, and it will to some extent when moderated by the human body, then the clock should be pretty accurate too.

 

[Deleted member 11852]

A watch that's normally on your wrist can perhaps be calibrated to keep time accurately at around body temperature. If the temperature remains constant, and it will to some extent when moderated by the human body, then the clock should be pretty accurate too.

Although it's stated that the typical body temperature is 98.6F in relaity your body temperature is almost never constant. It changes with age, the time of day, your metabolism (whether you're hungry or have just eaten) and how active you are. The only circumstance where your body temp is constant for a prolonged period is when you're dead!

 

[function]

Well, yeah.

Temperature of a watch strapped to you should be more constant than, say the clock in my car, which has to handle +40 to -10 degrees. Or in a device in my old house, which felt like it had to deal with temperatures of between ... +40 and -10 degrees.

Dead people are probably at their most constant temperature when they're six feet under. But then they can't tell the time because it's dark down there, so it doesn't matter.

 

[Albuquerque]

I love that kind of nitpicking and caveat.
You can try the software side at least.

Didn't even see this reply to the thread until today -- months later

Sorry to Blaz if I didn't communicate well; my intention was truly to be helpful. One of the few things I'm good at is enterprise virtualization, and time services are always a conversation point in any enterprise-level virtualization system. In my years, I've seen more than a few authoritative time servers installed as a guest VM, shortly followed by epic quantities of problems with systems that are sensitive to millisecond-resolution time variance.

Any application that requires a high level of time resolution must be tested and deployed in a very specific way and managed carefully when living as a VM. In many cases, a less mature IT organization should simply elect to deploy such apps in a physical device. For anyone playing along at home, you should ignore virtualization for authoritative time services or any extremely time-sensitive app.

 

[Blazkowicz]

Didn't even see this reply to the thread until today -- months later

Sorry to Blaz if I didn't communicate well; my intention was truly to be helpful.

I in turn perhaps didn't communicate that well, I mean that finding out about such issues when approfounding the subject is interesting and fun.