Hello @sersch,
Thanks for your thoughtful reply. I understand your position, you are just a customer of Siemens AG, a huge conglomerate and minor customer complaints probably are just a buzz down somewhere grass level that large behemoth hardly hears. I know, i’ve been same situation many times and had two find a workaround just to get job done like you.
Also those S7-300 and S7-400 seem to be already have legacy mark some sources I got searching and trying to get glimpse what kind of devices those are. So your luck getting them much attention to this is probably quite low.
Anyway. about NTP in general and time keeping a bit more. I write this not just for you, but perhaps someone else reading this thread also.
With time keeping always it depends greatly what kind of accuracy is needed.
-
SNTP synchronisation good enough, when distinct SNTP-clients do not have need much less than quarter minute time coordination difference issue to them. SNTP is been a while more a consumer and low accuracy time needing devices thing. It was created early 90’s by Prof. David L. Mills, same guy who created NTP protocol and started working network time sync problem from 1977, for then (90’s) low computing capability and power needs.
-
NTP is oldest still used original internet protocols because it was fortunately then really well designed. A proper time keeper and fills needs guide well with good time sources up to that already mentioned 10ms which it can fulfil even with moderate link speeds. When you are able to monitor it “ntpq -p” (I wish Teltonika would add ntpq tool to RutOS) or like you can see that it usually is even better just few milliseconds if good time sources are close enough and stable. NTP is good for general use data processing and services and syncing time over WAN interfaces too.
-
PTP (Precision Time Protocol, IEEE 1588) is newer and much more accurate than NTP. That is a thing that today where more than that 10ms accurate time needing industries should already be transitioning. PTP can keep good time sync even nanoseconds accuracy, with appropriate hardware. Most use of PTP is usually LAN only applicable.
-
And then when there are even more demanding environments like in physical sciences field which they need better than any of above protocols can provide. One of those is WR (CERN White Rabbit Project). I did not work any of that but was following interested when i knew guy who did tests and built Finnish NREN (FUNET) PTP WR network solution. Because at the time our physics department was considering particle accelerators (2 cyclotrons) next option instead having own time base when that would need to be upgraded. But I retired before that happened. Mikes in that pdf is Mittakeskus, the national metrology center and Metsähovi is a radio astronomy observatory that were primary users of that solution then 5 years ago.
OK, then just few more words about NTP usage, what I learned decades using it.
When you set up a site, which needs NTP sync. Don’t ever rely on just one time NTP server and nor two of them. It doesn’t matter if that time NTP servers time sources are GNSS or whatever.
You need alway three NTP services (site NTP servers or remote NTP) to make a proper synchronised time available for your NTP / SNTP clients.
Why?, you may ask?
Right, if you think about it a while only one NTP time source is clearly a single point of failure (SPoF) which is not good. Then OK, you tink I’ve have two of them and that will be good?
No, it isn’t. Why? Because if you think of it when you have two any kind of clocks how can you tell if which one of them is drifting?
The thing is there is no reliable way to know from two clocks which one is showing correct time and which one isn’t. Therefore you need three clocks as then you statistically very unlikely do not have two clocks drifting at same time and you can observer which one is drifting or broken.
Therefore always build a site with either three NTP servers, and they can be either local or remote as long as delay and jitter (variiation of delay) are within NTP protocol acceptable ranges.
Its a good working practise to add all three of your site NTP servers to each others as “peer” and any remote as “server” time sources. NTP protocol is quite smart, it will take care of rest. It monitor and follow which are at any time best stable time sources and decide to sync on these. It does not make your NTP kept time suddenly jump any direction. Any correction will be done either little by little slowing time if it needs to be done and same if it needs to catch up and was behind before. You don’t need to adjust any ways, just let those run and you are good when you have good multiple time sources with all of your NTP servers.
Also when you use remote NTP servers as time sources, do not configure all with exactly same time sources (other NTP servers, GNSS) if you can (are allowed and capable doing that), then you have much better change not to run in the situation that none of your site NTP servers have good time source(s) available.
If and when your site can access time sources from internet, you have multiple very likely decent NTP service pools available which using you get good time sources quite easily and it doesn’t cost you any more than the traffics you are paying using your site connection. Check www.ntp.org and their NTP pool project sources about what is available.
And please, do yourselves a favour do not run you own NTP servers in virtualised services in any kind of VM platform if you need or even appreciate in any ways good time with your systems. Local NTP-service is OK to run and keep time for that single system if it’s better than VM-agent (guest running daemon) time sync can provide. VM’s can work for a while, but may surprise you when the VM Host us under heavy load some later time.
Another thing with time to consider is what time your system should, if there is an option, select to run time internally? This writing is so long already that I do not go in to that now, but that is another time keeping related matter which can bite occasionally unaware.
Above is what I think applies most use cases NTP is used. Most important thing is not to just set one NTP-server and think that is all you need. It may be OK for a home users devices time source, but not much of anything else in my experience.
I had once a manager who thought I try to overdo this time thing when applying money upgrading time servers then recession time. But when asked back that how much he thinks would it cost if consequence of not enough accurate time for other time depending services were malfunctioning due NTP-service outages he understood and approved my request.
And I’m sorry anyone feeling bad, that I wrote so long post. But I did it because I good time keeping meaning and how NTP-services should be run are bit too little understood matter. Not that I everyone here this forum do not understand. Sure there are many, but as I see there are quite many with less experience they may think nah, I just set one NTP-server and be good. With that decision industrial and commercial environments you are most likely playing with luck more than you think you really are, if anything using that single device failure accumulate bit more than you did not think first. If that single NTP-server was not your decision you are bare network engineer, installer and not one designed it. Tell these things upwards who designs so that he/she can do educated guess and upgrade the thing when it is appropriate and would be needed. I know few examples when this was done and who I notified was later very happy I did this. He had not thought and understood this before I brought this to his attention. He was just folloiwing a customer request which mentioned in singular NTP-service ie. not asked many and redundant solution.
I think that’s enough for me. Again, apologies writhing so long post.
