Let's reduce our use of network time server synchronisation!

[E-waste]

18 hours ago, Vishera said:

Around 4 minutes.

Just reading that, it really doesn't seem that much, but that's quite a bit.  So in one year, your clock will drift by nearly 48+ minutes.  Even so, if you test this, by disabling the windows time service or turning off network sync in the date / time settings, it probably won't cause a single issue.

 

That is, if it's slowly drifting backward.  Shifting forward, 48 or possibly even 30 minutes might cause https connection issues, but you can always test this, to find iut just how far forward your clock can be, until issues begin.

 

Then make note of this, and just manually adjust periodically since you already figured the drift rate.

[BondiBlue]

On 1/7/2023 at 1:33 PM, Timpster said:

If we all gave just a small amount of thought as to how often we sync our clocks, and how we do so (offline vs network) I think it would make an absolutely MASSIVE difference to our energy use, from just this setting alone.

Can you please explain why you think NTP servers use a lot of power? There are countless other ways to easily reduce power consumption on a computer that would have a greater effect than disabling automatic time synchronization. 

[E-waste]

1 hour ago, BondiBlue said:

Can you please explain why you think NTP servers use a lot of power?

Sure and I supoort your curiosisty with my fascination of this topic.  After I answer your question, I'll list other ways to reduce our energy use regarding how we use computers.

 

NTP isn't needed for an overwhelming majority of computers to browse the web, check email, and play online multiplayer games.  NTP syncing is completely in our control and very simple to turn off.

 

Edit: There are 4,400 NTP pool servers for the ntp pool dot org group around the globe.  It doesn't take much to just ping for time, so this seems to me we're doing this more frequently than we need.

 

Here are other ways to reducr power use:

 

• NoScript, turning off / not loading the umpteen domaims on a website (even this one!) and not uploading various unknown bits of data, going through 20-40+ network hops each time.
• Domain / DNS filter, I probably shouldn't say exactly how much I block, but on Windows, I'll disclose I block the connection test, v10.events, msft dns, and many others that windows is programmed to connect to.  I feel this helps save power.  In fact I can't even connect to microsoft.com, I block quite a bit
• All the various MS Edge (a.ring and b.ring all the way to t.ring, and a few others, thus hopefully saving power, even though I don't prefer chromium code based software or browsers
• All Mozilla domains, including their forced Pocket service and shavar services, which I can't figure out how to disable
• For roku software (they won't appreciate this--but because they are 100% opaque in the data uploaded, that's what they'll get: no data) I block anythimg with roku in the domaim name, to the point that I can't update the TV.
• I block Netflix and Apple apps on the device from loading every bootup, and various image domains to try preventing images from loading when I can just read the names
• I run Pi-hole to do all of this and I cache my domain resolutions with Unbound
• I load my unbound dns cache file on every boot daily, and this reduces the energy needed to visit a webpage by getting the IP from the local cache.  My cache settings in unbound are set in the months, instead of the default of 175 seconds drfault w/o this, or unbounds default of 86400 seconds, one day.
• Lastly, I'm running my quad-core on just two cores, downclocked, undervolted EVERY SETTING in the uefi, some of them to their absolute minimum, unplugged all case fans except for gpu, cpu and one rear case fan.  Could probably unplug it too since it's cooler this time of year.

I'm not trying to brag, this is what I do, and I'd rather not think about how much server electricity use we could eliminate if everyone turn all this stuff off. and only updated their devices in bulk amounts of update, once or twice a year.

 

There is so much more we could ne doimg to save power absolutely.  We could hold on to slower, more power hungry hardware, and prevent the need or demand for more slave labor in both mineral mining and slave labor in manufacturing and all the plastic involved in packaging and shipping.

 

But it's most likeky unfeasible to expect ANYONE to go through these lengths in the attempt to reduce their electricity usage, and it's more likely thst someone will uncheck a box, so at lesst we have a better chance to make a difference.

[BondiBlue]

4 minutes ago, Timpster said:

Sure and I supoort your curiosisty with my fascination of this topic.  After I answer your question, I'll list other ways to reduce our energy use regarding how we use computers.

 

NTP isn't needed for an overwhelming majority of computers to browse the web, check email, and play online muktiplayer games.  NTP syncing is completely in our control and very simple to turn off.

This doesn't answer my question at all. You still haven't said why you believe NTP specifically is a large chunk of our global power consumption. You seem to think that using NTP less would somehow make a large dent in our energy usage, but why? 

 

6 minutes ago, Timpster said:

Lastly, I'm running my quad-core on just two cores, downclocked, undervolted EVERY SETTING in the uefi, some of them to their absolute minimum, unplugged all case fans except for gpu, cpu and one rear case fan.  Could probably unplug it too since it's cooler this time of year.

Umm, I'm sure you know this but the power you're saving by going to such lengths is basically nothing if you're talking about global power consumption (like you seem to be doing with your NTP claim). If you're doing this to save on your power bill then that's one thing, but for global energy savings it's nothing. 

[Lurick]

You didn't do much research did you? Windows, by default, syncs time ONCE a day or once a week to time.windows.com! This is a needless exercise in futility. Your Pihole VM/container/whatever uses more energy than all these DNS requests you're blocking, blocking doesn't mean it's not consuming any processing power, it's just going to null instead of being routed over the internet but they are still being processed by something. You're talking about saving a few watts of power over the course of weeks by blocking some requests locally. You're doing nothing in the grand scheme of things, the just a couple switches powering up in my lab consumes more power than you'd ever possibly save in a lifetime by doing this. Heck a single instance of windows updates probably nullify all your "savings" in one go.

 

Edit:

My statement about once a day syncs was incorrect, at least for Windows 11, it syncs once every 9 to 12 hours by default.

[Zando_]

59 minutes ago, BondiBlue said:

Umm, I'm sure you know this but the power you're saving by going to such lengths is basically nothing if you're talking about global power consumption (like you seem to be doing with your NTP claim). If you're doing this to save on your power bill then that's one thing, but for global energy savings it's nothing. 

To add on to this, said tweaking probably still results in a system that draws more power than modern low-TDP parts, while being slower to boot and thus more wasteful with power. Especially the GTX 960, Maxwell is not an efficient architecture.

[E-waste]

9 minutes ago, Zando_ said:

To add on to this, said tweaking probably still results in a system that draws more power than modern low-TDP parts

How about this: How much fuel, human sufferring, coal-burned in powering the nanufacturing facilities, plastic and other non-biodegradable packaging, and diesel and gasoline does it require to create a "lower power" comouter component.

 

It's going to be very difficult to convince me that my (before all these tweaks) my 48 watt idle system is going to eat up more power over time than it is to buy new hardware and thr energy input required, to have a system that draws 5 or 10 watts idle.

 

I don't like this idea, I just think it doesn't take into account the energy required to mine, manufacture. and ship. and then shio again to your business or home.  I've thought a lot about it too, this isn't something I just started considering.

 

So what about just tallying up how long a light is on (for fluorescents, it's actually better to keave them on, so the ballast and bulb don't go through as many rapid, frequent restarts) or however else we can save power.

 

What you're trying to convince me of, is that using 40 less watts a majority of the day, every day will make up for everything that goes into the manufacturing and all of new hardware.  Is there a study that can bear out these numbers and state this with evidence, it just doesn't seem accurate.

 

Of course, maybe I'm wrong, so what are your thoughts?

[Zando_]

4 minutes ago, Timpster said:

What you're trying to convince me of, is that using 40 less watts a majority of the day, every day will make up for everything that goes into the manufacturing and all of new hardware.  Is there a study that can bear out these numbers and state this with evidence, it just doesn't seem accurate.

You're very close to the point. The amount of power your PC pulls is completely inconsequential. I pointed out the inefficiency to highlight how you're not effectively sticking to your own proposed practices (it is incredibly easy to buy used parts that are exponentially more efficient, and also already manufactured so you are not necessitating the carbon/moral debt of a new piece of hardware being created). Same as the original call for disabling NTP, it would cause a lot of function, reliability, and efficiency issues and solve about 0.

 

If you find it satisfying to tweak old hardware to the lowest possible power draw, hell yeah. I have friends who enjoy seeing how efficient they can get their computers. It's just important to understand that that is not practical across the board for everyone, and certainly doesn't make any difference in global power consumption.

[Alex Atkin UK]

10 hours ago, Timpster said:

What software did you choose?  Also how frequently are you polling upstream, and what time servers have to choosen as upstream?

 

The NTP pool group is overworked and are needing more and more nodes to keep up with ever-increasing demands.  I'm hoping I can start the thought train with something as boring and dull as time syncing, and shed some light on how much we use it.  I'd like us to think if we need such down-to-the-minute accuracy, or could we be perfectly satisfied, with only syncing monthly, or even every time-shift for DST vs standard.

You're forgetting routers themselves, they sync every time you boot them as they rarely have battery backup.

 

I could imagine the huge uptick being from IoT devices which also probably wont hold their own time, its also important that every server and router on the Internet is keeping good time.  So simply stopping your PC from syncing as often is a drop in the ocean compared to where most of the requests are coming from.

 

I also think the impact varies, I'm sure my PC was once only a few minutes out and refused to work.  Plus as already mentioned there are the SSL certificates themselves, if they were updated AFTER the time your PC thinks it is, they wont work.

[Blue4130]

8 hours ago, Timpster said:

How about this: How much fuel, human sufferring, coal-burned in powering the nanufacturing facilities, plastic and other non-biodegradable packaging, and diesel and gasoline does it require to create a "lower power" comouter component.

 

It's going to be very difficult to convince me that my (before all these tweaks) my 48 watt idle system is going to eat up more power over time than it is to buy new hardware and thr energy input required, to have a system that draws 5 or 10 watts idle.

 

I don't like this idea, I just think it doesn't take into account the energy required to mine, manufacture. and ship. and then shio again to your business or home.  I've thought a lot about it too, this isn't something I just started considering.

 

So what about just tallying up how long a light is on (for fluorescents, it's actually better to keave them on, so the ballast and bulb don't go through as many rapid, frequent restarts) or however else we can save power.

 

What you're trying to convince me of, is that using 40 less watts a majority of the day, every day will make up for everything that goes into the manufacturing and all of new hardware.  Is there a study that can bear out these numbers and state this with evidence, it just doesn't seem accurate.

 

Of course, maybe I'm wrong, so what are your thoughts?

But why go after time servers specifically? They are really low powered and small servers compared to say, something like the Fugaku - Supercomputer that has 7,299,072 cores...

[Alex Atkin UK]

1 hour ago, Blue4130 said:

But why go after time servers specifically? They are really low powered and small servers compared to say, something like the Fugaku - Supercomputer that has 7,299,072 cores...

Yeah I think they forgot to read up on pool.ntp.org.

Quote

The pool.ntp.org project is a big virtual cluster of timeservers providing reliable easy to use NTP service for millions of clients.

The key words there are "virtual cluster".   Many of these are servers being used for other things that just happen to be hosting NTP as well, as anyone can become one.

[Blue4130]

40 minutes ago, Alex Atkin UK said:

Yeah I think they forgot to read up on pool.ntp.org.

The key words there are "virtual cluster".   Many of these are servers being used for other things that just happen to be hosting NTP as well, as anyone can become one.

It's just such an odd thing to try to save power on...

 

And for others that are curious. A generic non-virtualized NTP server is around 300w. The Fuji super computer that I listed is 29,900 kw. 29 MILLION watts. That is a lot of NTP servers...

[Alex Atkin UK]

26 minutes ago, Blue4130 said:

It's just such an odd thing to try to save power on...

 

And for others that are curious. A generic non-virtualized NTP server is around 300w. The Fuji super computer that I listed is 29,900 kw. 29 MILLION watts. That is a lot of NTP servers...

Especially as the bigger issue would be, how often is stuff like this happening and NOT being fixed?

[tikker]

19 hours ago, Timpster said:

The NTP pool group is overworked and are needing more and more nodes to keep up with ever-increasing demands.

  Is it really (genuine question that I'd be happy to see sources on) or do you just think that based on an estimate of how many daily queries it may get? I can't imagine a ping of "hey what's the time" is either a heavy load or power hungry compared to the copious amount of other things we try to do like streaming 8k video on YouTube.

19 hours ago, Timpster said:

I'm hoping I can start the thought train with something as boring and dull as time syncing, and shed some light on how much we use it.

The dull and boring is what keeps the wheels turning. You setting your clock backwards or forwards, for example, can be anything from not having synced in a long time, a benign attempt to extend a software trial to a malicious actor trying to trick you into using or accepting expired certificates. I have been greeted by the "your computer clock is wrong" pages in the past. Additionally, while for us humans it doesn't really matter if we are 1 minute early or late to a coffee appointment, under the hood it may matter a lot more to make sure networks remains secure.

19 hours ago, Timpster said:

Just reading that, it really doesn't seem that much, but that's quite a bit.  So in one year, your clock will drift by nearly 48+ minutes.  Even so, if you test this, by disabling the windows time service or turning off network sync in the date / time settings, it probably won't cause a single issue.

It causes a massive issue: you can no longer trust your clock that is anywhere from accurate moments after syncing to nearly an hour off within a year. How would you deal with your 4:15 appointment when you don't know when that is? The best you can do is assume that the drift is linear and on average drifts 4 minutes per month, but if you start accounting for that you may as well just sync the clock regularly, since then you are sure you have the correct time.

[Nayr438]

On 1/6/2023 at 4:02 PM, Timpster said:

My opinion is that 99% of all desktop PCs will function without issue, with zero network-based time synchronizationn.  If we instead manually set our clocks based on the ckock on a wall or oven or even our smartphones, that would be absokutely adequate to use our devices, without any clock-based issues ever.

18 hours ago, Timpster said:

NTP isn't needed for an overwhelming majority of computers to browse the web, check email, and play online multiplayer games.  NTP syncing is completely in our control and very simple to turn off.

 

Except many things depend on time.  It could be the difference of a certificate being valid and a 2Factor Code working. Your time being out sync could determine whether you can login to a service or whether a message actually gets delivered. it's also commonly used in Online Games and AntiCheat Systems for auto correction and validating data.

 

Devices tend to track time at different speeds, are periodically interrupted, suffer from interference, and gravity. In today's world we absolutely have a need for keeping time in sync and are the reason we rely on NTP Servers.

 

I have actually had a instance where I was temporarily locked out of one of my own services because my servers time was a minute behind, in that instance it was my two factor code not being valid due to the difference in time. Not saying that all services care, in fact some may not care at all and some may be super sensitive.

[wseaton]

On 1/8/2023 at 2:09 PM, E-waste said:

What software did you choose?  Also how frequently are you polling upstream, and what time servers have to choosen as upstream?

 

The NTP pool group is overworked and are needing more and more nodes to keep up with ever-increasing demands.  I'm hoping I can start the thought train with something as boring and dull as time syncing, and shed some light on how much we use it.  I'd like us to think if we need such down-to-the-minute accuracy, or could we be perfectly satisfied, with only syncing monthly, or even every time-shift for DST vs standard.

I use Meinberg. I just throw it on a domain controller, open the firewall port on the server and away you go. Windows based clients that are AD members take care of thrmselves in terms of time sync, but non AD machines or appliances, or even smart plugs aren't AD or even LDAP aware. But, I just point those devices at the server running Meinberg and I'm all set. Meinberg points at the NTP pool of my choice. Saves a whole lotta outbound egress that's for sure. 

 

Security cameras are the biggest one. They need fairly accurate and robust time stamps, but there's no way in hell I'm letting a bunch of chinese made cams running their own http servers have free roam outbound. I just point their NTP settings at the box running Meinberg, and done. They can't get at the internet and don't need to get at the internet.

 

To my frustration few Enterprise admins know about this trick, or that NTP servers are essentially free. 

 

 

[E-waste]

3 hours ago, wseaton said:

but non AD machines or appliances, or even smart plugs aren't AD or even LDAP aware. But, I just point those devices at the server running Meinberg and I'm all set. Meinberg points at the NTP pool of my choice. Saves a whole lotta outbound egress that's for sure.

This sounds way too good to be true! You're going to sit or stand at a keyboard and convince me that smart TVs and tablets, can be set to point at the NTP server of YOUR choosing?

 

Are you using firewall redirection or what?  This is amazing!

[E-waste]

On 1/8/2023 at 6:32 PM, Zando_ said:

it is incredibly easy to buy used parts that are exponentially more efficient, and also already manufactured so you are not necessitating the carbon/moral debt of a new piece of hardware being created

I really don't know how far yourvwilling to go down this thought exercise.

 

Here's my penny:

 

I agree 100% buying used is the absolute BEST thing we can do.  ZERO arguments therre.  But here's what I keep thinking about:

 

New hardware continues to get produced.  Clearly, you and I and all tech enthusiasts don't have a significant amount of control here.  But I can't help but think that well, if we keep buying used hardware, ... somewhere down the line, that somehow spurred or supported the ongoing demand for new hardware.

 

That new hardware quickly (in one or two years) becomes "not new" but this is not enough time to slow demand for more hardware.

 

What I truly want to do (and I probably won't nor will anyone else be able to) is slow demand for new hardware production.  I guess that's a fight I'll never win.  And the only reasons I want to do that is:

 

- Slave labor

 

- Electronic waste / recycling.

 

I'd love to live in a world, or help encourage a future population of people in this world, to have any hardware we want, if we would just recycle and not abuse others to collect and manufacture it.

 

So until we get it all staightened out, I'd really like to encourage everyone to slow down on technology aquisition for just a while, until that's a reality.

 

If we all just buy used, well I get the impression that still it's not enough and many hundreds of millions of people are still buying new hardware, completely oblivious to the attrocities that allowed to to be sold.  Oh if we could JUST get the word out a little bit more I think we could get closer to a solution.

[Zando_]

10 minutes ago, E-waste said:

But I can't help but think that well, if we keep buying used hardware, ... somewhere down the line, that somehow spurred or supported the ongoing demand for new hardware.

Been on this forum for a while now. In most cases people just... upgrade their PC. Then maybe they sell the old stuff, maybe they give it away, maybe they repurpose it, but it has no real bearing on buying the new hardware. In my experience it's relatively uncommon that someone gets a new PC because they can sell their old hardware to make up most of the budget. Been the same in the case of hardware I bought in-person, people had already upgraded, were selling it off because the extra cash is nice, it had no bearing on their decision to upgrade.