Re: [ntp:questions] Is dispersion jitter in all situations
On 5 Jan, 22:20, David Woolley da...@ex.djwhome.demon.invalid wrote: B wrote: I want to know the accuracy on a certain NTP-server at stratum 3. It is easy to calulate the absolute error bounds that wont be exceeded with this equation OFFSET +/- [DELTA/2 + DISPERSION]. This will in my case be OFFSET +/- 4 seconds, but I need to know more precise, ie an indicator of In that case there is something wrong in your configuration! ntpd will start ignoring servers if their root distance is rather less than this and it would be an unusual system where the leaf nodes go so long between polls that they can accumulate the additional dispersion needed to reach 4 seconds. Oh, sorry for my typo. In my case the absolute error bounds is -0,72 +/- 3,9 ms, not seconds! There is almost a dedicated network with optical fiber and no(very little) asymmetris. There is a huge network which got three GPS-clocks as stratum 0, reference clocks for the primary servers. Below you can see the selected servers, all in [milliseconds]. Stratum 2 ref clock st when poll reach delay offsetdisp .GPS. 15564 377 0.70.02 0.1 Stratum 3 ref clock st when poll reach delay offsetdisp xx.xx.x 2 154 256 377 5.7 -0.74 0.6 For my logging application and master thesis report I need to know a more narrow interval, and I know it is more accurate than the absolute error bounds [-4.62, 3.18]. So a smaller interval of expected time is desirable. Unruh you are probably right, but aren't you talking about synchronization distance(sometimes called root distance), DELAY/2 + DISPERSION, as a conservative estimate of the error. I thinking if it is ok to use OFFSET with DISPERSION as the uncertainty about that value. If only DISPERSION is a conservative value for that too, at least I got a narrower interval of [-0.72 - 0.7, -0.72 + 0.7]=[-1.42, -0,02] milliseconds. I really appriciate your help! // B ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
Dispersion is, the diff to stratum 0 (UTC). Root dispersion yes, but there is also a peer dispersion. Can you explain what you mean by the phrase expecting time? I want to know the accuracy on a certain NTP-server at stratum 3. It is easy to calulate the absolute error bounds that wont be exceeded with this equation OFFSET +/- [DELTA/2 + DISPERSION]. This will in my case be OFFSET +/- 4 seconds, but I need to know more precise, ie an indicator of expecting time. Becouse the distribution isn't known within these interval, I can't say that the time at stratum 3 will be within 2 ms relative UTC at 95% of the time. David L Mills wrote in an old thread: Use ntpq and the rv billboard for the rootdelay, rootdispersion and jitter displays. Note the jitter display, which includes both peer jitter and selection jitter, is probably the best indicator of expected time quality. Read this as follows: the best estimate of the server time is the offset in the rv display, with jitter as the uncertainty about that value. Absolute error bounds is within this interval [OFFSET - DELTA/2 - DISPERSION, OFFSET+DELTA/2 + DISPERSION] DISPERSION, in this case PEER.DISPERSION is defined in RFC-1305, page 102 as the maximum error in OFFSET and the maximum error in ROUNDTRIP DELAY. PEER.DISPERION is the maximum error in the interval. If PEER.DISPERSION is bigger than JITTER. I will able to use OFFSET +/- PEER.DISPERSION as anestimate of the server time by analogy with what David L Mills wrote: the best estimate of the server time is the offset in the rv display, with jitter as the uncertainty about that value. Hope you understand my question Best regards / B ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
On 2010-01-05, B berti...@gmail.com wrote: Dispersion is, the diff to stratum 0 (UTC). Root dispersion yes, but there is also a peer dispersion. Can you explain what you mean by the phrase expecting time? I want to know the accuracy on a certain NTP-server at stratum 3. It is easy to calulate the absolute error bounds that wont be exceeded with this equation OFFSET +/- [DELTA/2 + DISPERSION]. This will in my case be OFFSET +/- 4 seconds, but I need to know more precise, ie an indicator of expecting time. Becouse the distribution isn't known within these interval, I can't say that the time at stratum 3 will be within 2 ms relative UTC at 95% of the time. David L Mills wrote in an old thread: Use ntpq and the rv billboard for the rootdelay, rootdispersion and jitter displays. Note the jitter display, which includes both peer jitter and selection jitter, is probably the best indicator of expected time quality. Read this as follows: the best estimate of the server time is the offset in the rv display, with jitter as the uncertainty about that value. Absolute error bounds is within this interval [OFFSET - DELTA/2 - DISPERSION, OFFSET+DELTA/2 + DISPERSION] DISPERSION, in this case PEER.DISPERSION is defined in RFC-1305, page 102 as the maximum error in OFFSET and the maximum error in ROUNDTRIP DELAY. PEER.DISPERION is the maximum error in the interval. If PEER.DISPERSION is bigger than JITTER. I will able to use OFFSET +/- PEER.DISPERSION as anestimate of the server time by analogy with what David L Mills wrote: the best estimate of the server time is the offset in the rv display, with jitter as the uncertainty about that value. No. The dispersion is a very very conservative ( read large) estimate of the error in the time. Unless you are really really unlucky ( eg, each outbound packet takes 1usec to get to the stratum 2 server, and 1 second to get back) the dispersion is going to be very much larger than the actual difference between your time and UTC. However, the only way you can really know is to get a GPS with PPS, hook it up to your machine and measure the difference in time between the Stratum 3 set clock and the GPS time. Of course, if you have that gps, you might as well use it to set the time and you then have a stratum 1 time. Assuming that you do not have such a weird, consistantly assymetric link to your server, then the jitter is probably a better estimate of the time difference from UTC. In the science literature, this is called the difference between random errors and systematic errors. The latter are always far far harder to estimate than the former. Hope you understand my question Best regards / B ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
Bill Unruh writes: However, the only way you can really know is to get a GPS with PPS, hook it up to your machine and measure the difference in time between the Stratum 3 set clock and the GPS time. I think that what he wants is the expected time (he has that: the time Ntpd reports) and a 95% confidence interval for it. That is, he wants to be able to say The time is within +- 4.2usec of 2300UTC with 95% confidence. -- John Hasler jhas...@newsguy.com Dancing Horse Hill Elmwood, WI USA ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
On 2010-01-05, John Hasler jhas...@newsguy.com wrote: Bill Unruh writes: However, the only way you can really know is to get a GPS with PPS, hook it up to your machine and measure the difference in time between the Stratum 3 set clock and the GPS time. I think that what he wants is the expected time (he has that: the time Ntpd reports) and a 95% confidence interval for it. That is, he wants to be able to say The time is within +- 4.2usec of 2300UTC with 95% confidence. Unfortunately he cannot get that. the random part he can get (jitter) but the systematic part he cannot. The dispersion is a very very conservative estimate of the systematic part, but it is in almost all situations far far larger than the true 95% confidence interval. As I said the only way to estimate the systematics is by having a local gps pps clock to compare it with, but that is obviously overkill, since then you would just use the gps to give the time. for example, I have a machine with a gps pps and with another source being a stratum 1 server 2000km away. The round trip time is 45ms, which would give a dispersion component of 22.5ms. But actually the offset of that source from the local gps time is only about +-.1ms.Thus the estimate -- 1/2 the roundtrip -- of the systematic error is out by about a factor of 200. Thus the estimate -- 1/2 the roundtrip -- of the systematic error is out by about a factor of 200. As I said, it is possible that all outgoing ntp requests go via a 1Gb ethernet, and all return packets go via a 300baud modem. In that case the estimate of 1/2 the round trip would be a good estimate of the systematic error. Unfortunately there is absolutely nothing ntp can do on its own to figure that out. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
B wrote: I want to know the accuracy on a certain NTP-server at stratum 3. It is easy to calulate the absolute error bounds that wont be exceeded with this equation OFFSET +/- [DELTA/2 + DISPERSION]. This will in my case be OFFSET +/- 4 seconds, but I need to know more precise, ie an indicator of In that case there is something wrong in your configuration! ntpd will start ignoring servers if their root distance is rather less than this and it would be an unusual system where the leaf nodes go so long between polls that they can accumulate the additional dispersion needed to reach 4 seconds. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
In article slrnhk7469.kji.un...@wormhole.physics.ubc.ca, unruh un...@wormhole.physics.ubc.ca writes: As I said, it is possible that all outgoing ntp requests go via a 1Gb ethernet, and all return packets go via a 300baud modem. In that case the estimate of 1/2 the round trip would be a good estimate of the systematic error. Unfortunately there is absolutely nothing ntp can do on its own to figure that out. A more likely cause for asymmetry is queueing delays on home DSL links. Nasty. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] Is dispersion jitter in all situations
Hi, this really concerns me! An indicator of expecting time is important and I know about the error bounds, but I want to use a value(jitter) as an indicator of the expecting time relative offset. The jitter isn't introduced before NTPv4 and where I am doing my master thesis they are using NTPv3(RFC-1305). Is it possible to use dispersion relative offset as an indicator of expecting time? My idea, peer.dispersion represents the maximum error in offset and maximum error of half the roundtrip delay. If dispersion is bigger than jitter, ie jitter is bounded by dispersion, then dispersion could be used as an indicator of expecting time relative offset. Any guidance in this is appreciated? Best regards ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
On 1/4/2010 7:30 AM, B wrote: An indicator of expecting time is important and I know about the error bounds, but I want to use a value(jitter) as an indicator of the expecting time relative offset. The jitter isn't introduced before NTPv4 and where I am doing my master thesis they are using NTPv3(RFC-1305). Is it possible to use dispersion relative offset as an indicator of expecting time? My idea, peer.dispersion represents the maximum error in offset and maximum error of half the roundtrip delay. If dispersion is bigger than jitter, ie jitter is bounded by dispersion, then dispersion could be used as an indicator of expecting time relative offset. AFAICT Dispersion is, the diff to stratum 0 (UTC). Offset is the diff to combined offset from the selected truechimer peers reference times. Jitter is relative to that combined peer offset, not to strat 0 (UTC). -- E-Mail Sent to this address blackl...@anitech-systems.com will be added to the BlackLists. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
B wrote: Hi, this really concerns me! An indicator of expecting time is important and I know about the error bounds, but I want to use a value(jitter) as an indicator of the expecting time relative offset. The jitter isn't introduced before NTPv4 and where I am doing my master thesis they are using NTPv3(RFC-1305). Is it possible to use dispersion relative offset as an indicator of expecting time? Can you explain what you mean by the phrase expecting time? ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
B wrote: Hi, this really concerns me! An indicator of expecting time is important and I know about the error bounds, but I want to use a value(jitter) as an indicator of the expecting time relative offset. The jitter isn't introduced before NTPv4 and where I am doing my master thesis they are using NTPv3(RFC-1305). You might want to ask Uppsala admins why they are using such an old version. It hasn't been supported for years and most Unix O/S's are now shipping V4 and have done so for years. You could also install your own version of NTP on your own system so that you can conduct your own experiments. Is it possible to use dispersion relative offset as an indicator of expecting time? My idea, peer.dispersion represents the maximum error in offset and maximum error of half the roundtrip delay. If dispersion is bigger than jitter, ie jitter is bounded by dispersion, then dispersion could be used as an indicator of expecting time relative offset. Jitter is defined in Section 4 of the draft NTP V4 RFC as follows: The jitter (psi) is defined as the root-mean-square (RMS) average of the most recent offset differences, represents the nominal error in estimating the offset. Dispersion however is also defined in the same Section 4 as follows: The dispersion (epsilon) represents the maximum error inherent in the measurement. These are very different from each other. The first gives you information about how much the offset is varying when receiving NTP packets while dispersion is a measure of your ability to measure a value but gives no indication of the current value of the jitter. What exactly are you trying to accomplish here? Danny -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions