[ntp:questions] Routers and NTP Timing loops
I had an interesting experience with a Cisco router as ntp master 9 using itself as an NTP source and it's neighboring switch as the only other NTP source. The switch had the exact same configuration right back at the router!. So what happens? Is NTP smart enough to know that this is a bad idea? If not, what goofy things happen if you establish an NTP timing loop? I know that if this were frequency references and stratum 3 clocks, they would ultimately peg at the end of their frequency range (not good) but I don't know what happens from a time perspective. I had some interesting date-time issues and this is a potential root cause. Just looking to see if anyone has had similar experience and have done some in-depth research in this area. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Reference clock driver for /dev/rtc
Hi, -Original Message- From: unruh [mailto:un...@wormhole.physics.ubc.ca] Sent: Tuesday, June 15, 2010 7:15 PM To: questions@lists.ntp.org Subject: Re: [ntp:questions] Reference clock driver for /dev/rtc On 2010-06-15, Krejci, Pavel pavel.kre...@siemens-enterprise.com wrote: Hi, since I cannot use kvm-clock as the clock source (older guest kernel) I am using pit as the clock source. According to my tests this is the most stable clock source among tsc,hpet but still can drift. Since the qemu keeps the /dev/rtc perfectly synchronized with the Host's system time it is a good time source for the ntpd on the guest. The host itself is then sychronized via NTP with the external time server. I don't know any other way how to read the system time from the Host, please offer if you have some. I do not understand. If you driver can read the rtc, it can read the system clock instead. I am not reading the Host's /dev/rtc. I am reading the Guest's /dev/rtc, which is synchronized with the Host's system clock. And virtual systems are terrible things to use ntpd on. ntpd cannot control something where the clock varies by more than 500PPM, and virtual systems, since they are shut down for variable lengths of time by the host, tend to have terrible clocks. Yes the clocks like hpet or tsc are drifting very very much and the ntpd cannot improve the resulting stability. But the pit keeps quite well. With additional ntpd the resulting long period clock stability is good - no exact measures done yet... The rtc can only be read in one second chunks. This does not matter, some radio clocks allow the same. The only disadvantage is that when the step time back must be done on the Host, the /dev/rtc gets stuck (it is a monotonic clock) and the qemu must be restarted. rtc is not a monotonic clock. It can be set to whatever time you want. Unless your hardware is different than what I am imagining. This comes from the implementation of the /dev/rtc by the qemu. I haven't investigated why. Regards Pavel Regards Pavel -Original Message- From: unruh [mailto:un...@wormhole.physics.ubc.ca] Sent: Tuesday, June 15, 2010 2:23 AM To: questions@lists.ntp.org Subject: Re: [ntp:questions] Reference clock driver for /dev/rtc On 2010-06-14, Krejci, Pavel pavel.kre...@siemens-enterprise.com wrote: Hello, I have written the reference clock driver for /dev/rtc on Linux. We use it to synchronize the guest Linux system running in the qemu with the Host clock. If this is useful to someone else I would like to contribute to the NTP project. How should I proceed? Why would you want to do that? The rtc is almost certainly worse than the system clock. Why not have the guest just read the host's system clock, rather than the rtc. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Routers and NTP Timing loops
Jacobs, Kevin J. kjac...@mitre.org wrote in message news:70189427bd8ce046b781f28d29c21aee0c9bd78...@imcmbx4.mitre.org... I had an interesting experience with a Cisco router as ntp master 9 using itself as an NTP source and it's neighboring switch as the only other NTP source. The switch had the exact same configuration right back at the router!. So what happens? Is NTP smart enough to know that this is a bad idea? If not, what goofy things happen if you establish an NTP timing loop? I know that if this were frequency references and stratum 3 clocks, they would ultimately peg at the end of their frequency range (not good) but I don't know what happens from a time perspective. I had some interesting date-time issues and this is a potential root cause. Just looking to see if anyone has had similar experience and have done some in-depth research in this area. The reference implementation gets a refid with every association to prevent exactly this. If the refid of a server is recognised as that of the client itself, the server is not considered a viable time source. I don't think it breaks cycles of three or more machines. Groetjes, Maarten Wiltink ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
2010/6/15 unruh un...@wormhole.physics.ubc.ca On 2010-06-14, Marcelo Pimenta marcelopiment...@gmail.com wrote: Hi everybody!! My question is about Time Accuracy of NTP/SNTP protocol. I want to know if is possible to have precision of 1ms(could be 1 ms?) using SNTP in a network with 50 hosts(is the same with 100 hosts?) using only swithes, no routing. All these hosts are syncronized with a Meinberg GPS. The traffic is low(~25%). Is it possible to have the same time accuracy of IRIG-B for example? I think you mean accuracy of 1ms. Yes. it can. On a local network, 100usec is easily possible ( and depending on the routers, it can be better than that). Yes, you are right, accuracy of 1ms. On a local network 100usec?? Even if we use only switches(no routers), how is that possible if I have 4 types of Latency increasing about 80us? Algorithms to compensate the delay? Even in SNTP? Anyway, the question is about 1 ms. To get 100usec I'll use PTP. The internal clock in my devices(hosts) starts with the crystal frequence. A typical offset or frequence tolerance of the crystal is less then 1ms/minute. The resolution is about 0.9ms and the devices acts as unicast clients, asking GPS every minute. Why? The whole purpose of ntp is to discipline your clock so that it keeps time much better than that. The idea to have this is not overload the network asking time every second to keep my accurancy in 1ms in 100% of time. And the second reason is about how many ms your clock will be wrong in 59 sec without a frame to discipline my clock again? PCs clock for example is not good, that's why I need a very good crystal. Not sure what you mean by the resolution is about .9ms What is your device? I mean that my device can show me variations of 0.9ms, is the smallest possible increase of time the clock model allows. I work with protection IED(intelligent electronic device), is a term used in the electric power industry to describe microprocessor-based controllers of power system equipment. If something happens in 458ms, I will get exacly 458ms. Some equipment have so I can trust in variation that I can capture. Is it enough? The Time Accurancy could be better if I could ask GPS twice a minute? A broadcast implementation with GPS sending time every minute could help? It depends on what you want the time for. I need accurancy at least of 1ms in 100% of time. I know that these kind of issue have many variables but the main question is: Is possible to have precision of 1ms(could be 1ms) with SNTP? Yes, depending on the competence of the writer of the SNTP software. So, the problem is not in the standard but into implementation of SNTP to not loose time to correct the clock and something like that? Thank you very much ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
Marcelo Pimenta marcelopiment...@gmail.com wrote: The idea to have this is not overload the network asking time every second to keep my accurancy in 1ms in 100% of time. And the second reason is about how many ms your clock will be wrong in 59 sec without a frame to discipline my clock again? PCs clock for example is not good, that's why I need a very good crystal. Why are you causing your own problems by sticking to SNTP instead of using NTP which was designed to overcome the problems you fear? ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
2010/6/16 Rob nom...@example.com Marcelo Pimenta marcelopiment...@gmail.com wrote: The idea to have this is not overload the network asking time every second to keep my accurancy in 1ms in 100% of time. And the second reason is about how many ms your clock will be wrong in 59 sec without a frame to discipline my clock again? PCs clock for example is not good, that's why I need a very good crystal. Why are you causing your own problems by sticking to SNTP instead of using NTP which was designed to overcome the problems you fear? Rob, my understading about the use of SNTP and NTP is: while SNTP provides time synchronization within *one *network, NTP allows a global time synchronization on the internet. SNTP provides the current time, the current number of leap seconds and the warning flags marking the intriduction of a leap second correction. The NTP algorithm is much more complicated than the SNTP algorithm. NTP normally uses multiple time servers to verify the time and then controls the slew rate of the device. The algorithm determines if the values are accurate using several methods including fudge factors and identifying time servers that don't agree with the other time servers. It then speeds up or slows down the device drift rate so device time is always correct and there won't be any subsequent time jumps after the initial correction. SNTP usually uses just one Ethernet Time Server to calculate the time and then it jumps the system time to the calculated time. It can, however, have back-up Ethernet Time Servers in case one is not avaiable(not two at the same time). In my case, I have* only one* network. My Time Server is not a machine, is a meinberg GPS. In my point of view, if my source time were machines, maybe NTP could be better to find a middle line between all these machines used as time servers. But if I am using a very good and reliable GPS(Meinberg) with a lot of satellites giving it the correct time, and it's pluged directly in my switch, I think that in this case, NTP will not make any difference. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Routers and NTP Timing loops
Maarten Wiltink wrote: I don't think it breaks cycles of three or more machines. However, unless you abuse the local clock, which many packagings of ntpd do, the stratum mechanism will cut you off. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
Marcelo Pimenta wrote: Rob, my understading about the use of SNTP and NTP is: while SNTP provides time synchronization within *one *network, NTP allows a global time You are confusing it with timed. SNTP also expects to use global time, and NTP can be used with an arbitrary timebase, providing all stratum 0s have the same arbitrary time (i.e., typically there is only one). synchronization on the internet. SNTP provides the current time, the current number of leap seconds and the warning flags marking the intriduction of a leap second correction. The NTP algorithm is much more complicated than the SNTP algorithm. NTP normally uses multiple time servers to verify the time and then controls the There is no SNTP algorithm, beyond basic validity checking. SNTP covers any use of NTP wire formats that falls short of being a compliant NTP implementation. SNTP implementations could still use PLL clock discipline code, or could use linear regression solutions for offset and rate. slew rate of the device. The algorithm determines if the values are accurate using several methods including fudge factors and identifying time servers that don't agree with the other time servers. It then speeds up or slows down the device drift rate so device time is always correct and there won't be any subsequent time jumps after the initial correction. SNTP usually uses just one Ethernet Time Server to calculate the time and then it jumps the system time to the calculated time. It can, however, One issue that has maybe not been stressed is that this strategy makes you extremely vulnerable to scheduling delays, e.g. you would likely never get a 1ms error bound on Windows if you used this strategy. Early W32Time implementations, and current ones in out of the box configurations are SNTP implementations, but often use internet sources. They do violate the specification by allowing multiple SNTP strata. have back-up Ethernet Time Servers in case one is not avaiable(not two at the same time). In my case, I have* only one* network. My Time Server is not a machine, is a meinberg GPS. In my point of view, if my source time were machines, maybe NTP could be better to find a middle line between all these machines used as time servers. But if I am using a very good and reliable GPS(Meinberg) with a lot of satellites giving it the correct time, and it's pluged directly in my switch, I think that in this case, NTP will not make any difference. I think you are being asked why you don't just use a standard ntpd for the platform, and also how you know that your total system meets the scheduling constraints for 1ms accuracy with SNTP. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
Marcelo Pimenta marcelopiment...@gmail.com wrote: In my case, I have* only one* network. My Time Server is not a machine, is a meinberg GPS. In my point of view, if my source time were machines, maybe NTP could be better to find a middle line between all these machines used as time servers. But if I am using a very good and reliable GPS(Meinberg) with a lot of satellites giving it the correct time, and it's pluged directly in my switch, I think that in this case, NTP will not make any difference. If that is your opinion, you should not complain about network latency and clock rate errors. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
Marcelo Pimenta marcelopiment...@gmail.com wrote in message news:aanlktilq6m8apeoasibr-o8mhwifqkfv9xyf6mudr...@mail.gmail.com... [...] The NTP algorithm is much more complicated than the SNTP algorithm. The short, short version: there is no SNTP algorithm. SNTP is NTP _without_ the algorithms. Using NTP means continuously adjusting the speed of your clock so it tracks real time as best you can make it, while SNTP is simply asking what time [they think] it is. Groetjes, Maarten Wiltink ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
Marcelo Pimenta marcelopiment...@gmail.com wrote in message news:aanlktilaoduniqjgpigohpzvjcv_zmsw_tr7naj6b...@mail.gmail.com... [...] accuracy of 1ms. On a local network 100usec?? Even if we use only switches(no routers), how is that possible if I have 4 types of Latency increasing about 80us? The latency is corrected for. A query-response cycle is a back-and-forth exchange. The assumption is that the latency for the response (going forth) is equal to the latency for the response (going back), and both are half of the total latency, which you do know. So you estimate that the response was sent half the total latency time before the moment you received it. This often works well, and sometimes not so well. Domestic ADSL with a saturated upstream and downstream capacity left is one common case where it works not so well. [...] I need accurancy at least of 1ms in 100% of time. That's a *very* tall order. Groetjes, Maarten Wiltink ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
