Question about authority clock skew
Hello everyone, how are you? I was browsing the "Consensus health" page and something let me curious... What is the importance of the clock skew in the authorities with the resolution of microseconds? Picture: 2020/04/13 - 13:37 UTC Also, how the authorities compare their clocks? Using ntp daemon for example? I'm asking this here because I work in a time and frequency reference lab and have two NTP (stratum 1, connected directly to UTC(LRTE) timescale. And a stratum 2 that is discilplinned with the stratum 1 timeserver). So considering the microssencond accuracy did the keepers of these authorities care about the time source? Or any time source is ok? Best regards and keep safe! Luiz
Hi Luiz,
On 13. Apr 2020, at 15:41, torjoy <south_america_bridges@protonmail.com> wrote: I was browsing the "Consensus health" page and something let me curious... What is the importance of the clock skew in the authorities with the resolution of microseconds? <Imagem colada-Mon Apr 13 2020 103833 GMT-0300 (Horário Padrão de Brasília).png> Picture: 2020/04/13 - 13:37 UTC
Also, how the authorities compare their clocks? Using ntp daemon for example? I'm asking this here because I work in a time and frequency reference lab and have two NTP (stratum 1, connected directly to UTC(LRTE) timescale. And a stratum 2 that is discilplinned with the stratum 1 timeserver). So considering the microssencond accuracy did the keepers of these authorities care about the time source? Or any time source is ok?
I operate the directory authority gabelmoo. We do not synchronize our clocks amongst ourselves and accuracy of +/- a few seconds is really not important. Every operator does their best to keep the time roughly correct individually, for example gabelmoo uses a site-local NTP server that gets physical time from the German government's time broadcasting service. The page you're referring to just shows the skew with a lot more digits than can actually be accurately measured. The reason to keep clock skew in check is that if the time differs by more than a few seconds, the voting process can get impacted which in the past has led to consensuses not being created even though enough directory authorities would theoretically be ready for it. Also some directory authorities had some historical trouble with keeping an accurate time due to virtual machine trouble - this was worrisome for relay operators, because they would get (wrong) warnings about a wrong time in their logfile. Hope that helps Sebastian
Hi Sebastian, Good to know! This is just a curiosity that I have. Did you know if some cryptographic operations uses high resolution like this (microssecond for example) to match or no somethings like certificates on our case (TOR)? Also, another thing that i'm always asking myself is if here in south america isn't interesting that we have some authority. What is the "actual load" and problems? I see here most of relays haven't the "real" bandwidth that they can really deliver... Is this a measurement problem caused by distance of the authorities? I have here on Brazil three bridges and one relay that i'm operating today. Thank you for the answer!! Luiz Sent with ProtonMail Secure Email. ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ Em Segunda-feira, 13 de Abril de 2020 às 19:30, Sebastian Hahn <sebastian@torproject.org> escreveu:
Hi Luiz,
On 13. Apr 2020, at 15:41, torjoy south_america_bridges@protonmail.com wrote: I was browsing the "Consensus health" page and something let me curious... What is the importance of the clock skew in the authorities with the resolution of microseconds? <Imagem colada-Mon Apr 13 2020 103833 GMT-0300 (Horário Padrão de Brasília).png> Picture: 2020/04/13 - 13:37 UTC Also, how the authorities compare their clocks? Using ntp daemon for example? I'm asking this here because I work in a time and frequency reference lab and have two NTP (stratum 1, connected directly to UTC(LRTE) timescale. And a stratum 2 that is discilplinned with the stratum 1 timeserver). So considering the microssencond accuracy did the keepers of these authorities care about the time source? Or any time source is ok?
I operate the directory authority gabelmoo. We do not synchronize our clocks amongst ourselves and accuracy of +/- a few seconds is really not important. Every operator does their best to keep the time roughly correct individually, for example gabelmoo uses a site-local NTP server that gets physical time from the German government's time broadcasting service.
The page you're referring to just shows the skew with a lot more digits than can actually be accurately measured. The reason to keep clock skew in check is that if the time differs by more than a few seconds, the voting process can get impacted which in the past has led to consensuses not being created even though enough directory authorities would theoretically be ready for it. Also some directory authorities had some historical trouble with keeping an accurate time due to virtual machine trouble - this was worrisome for relay operators, because they would get (wrong) warnings about a wrong time in their logfile.
Hope that helps Sebastian
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
On Mon, Apr 13, 2020 at 01:41:41PM +0000, torjoy wrote:
I was browsing the "Consensus health" page and something let me curious... What is the importance of the clock skew in the authorities with the resolution of microseconds?
Sebastian's answer is exactly right. I've just opened https://bugs.torproject.org/33896 because you're right, there's no way the level of precision that consensus-health reports is at all accurate.
So considering the microssencond accuracy
Using the definitions that "precision" is how many digits you're providing, and "accuracy" is how right you are, I'd say that we're giving you microsecond precision but not microsecond accuracy. :) Thanks, --Roger
Thank you for the clarifications, Roger! So the "how right you are" is not too important for all the TOR network? Luiz Sent with ProtonMail Secure Email. ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ Em Segunda-feira, 13 de Abril de 2020 às 20:34, Roger Dingledine <arma@torproject.org> escreveu:
On Mon, Apr 13, 2020 at 01:41:41PM +0000, torjoy wrote:
I was browsing the "Consensus health" page and something let me curious... What is the importance of the clock skew in the authorities with the resolution of microseconds?
Sebastian's answer is exactly right.
I've just opened https://bugs.torproject.org/33896 because you're right, there's no way the level of precision that consensus-health reports is at all accurate.
So considering the microssencond accuracy
Using the definitions that "precision" is how many digits you're providing, and "accuracy" is how right you are, I'd say that we're giving you microsecond precision but not microsecond accuracy. :)
Thanks, --Roger
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
On 4/14/20 1:34 AM, Roger Dingledine wrote:
Using the definitions that "precision" is how many digits you're providing, and "accuracy" is how right you are, I'd say that we're giving you microsecond precision but not microsecond accuracy. :) Hehe, the first thing I was teached during my study was to not promise more accuracy than actually given ;-)
-- Toralf
participants (4)
-
Roger Dingledine -
Sebastian Hahn -
Toralf Förster -
torjoy