A slightly disheartening tale of NTP's possible future: <http://www.informationweek.com/it-life/ntps-fate-hinges-on-father-time/d/d-id/1319432>. (via MetaFilter, which has a few decent comments on the matter: <https://www.metafilter.com/147900/NTP-or-Not-NTP-That-is-the-question>) cheers, Stewart
This is an important topic. "Keeping time in sync" is a cornerstone to technology from gps to ssl, from hft to big data. I was sent back to school recently when I saw "chrony" and not ntp on my centos 7 box. I'm not sold yet. And i've not finished reading the horrible documentation for it. " The NTP time stamp is part of how equities firms show that trades took place when they say they did" - False. Ntp is not precise enough. My finance client's use ptp not ntp. David On Sat, Mar 14, 2015 at 4:25 PM, Stewart C. Russell <scruss@gmail.com> wrote:
A slightly disheartening tale of NTP's possible future: <http://www.informationweek.com/it-life/ntps-fate-hinges-on-father-time/d/d-id/1319432> <http://www.informationweek.com/it-life/ntps-fate-hinges-on-father-time/d/d-id/1319432> . (via MetaFilter, which has a few decent comments on the matter: <https://www.metafilter.com/147900/NTP-or-Not-NTP-That-is-the-question> <https://www.metafilter.com/147900/NTP-or-Not-NTP-That-is-the-question>)
cheers, Stewart
--- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
On 2015-03-14 10:09 PM, David Thornton wrote:
"Keeping time in sync" is a cornerstone to technology from gps to ssl, from hft to big data. Absolutely. At the solar plant I audited the other week, I was amused to see that the great big huge expensive grid protection and control box was controlled via GPS sync. It's the cheapest way to maintain grid frequency at the far end of the grid. And I mean /far/; 5 hours west of Thunder Bay …
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete. cheers, Stewart
On 03/15/2015 10:36 AM, Stewart C. Russell wrote:
On 2015-03-14 10:09 PM, David Thornton wrote:
"Keeping time in sync" is a cornerstone to technology from gps to ssl, from hft to big data. Absolutely. At the solar plant I audited the other week, I was amused to see that the great big huge expensive grid protection and control box was controlled via GPS sync. It's the cheapest way to maintain grid frequency at the far end of the grid. And I mean /far/; 5 hours west of Thunder Bay
Out in the middle of the Pacific??? The left coast is only 3 hours behind. A wavelength at 60 Hz is 5000 Km, though a bit shorter in power lines. With 3 phase power, there are 6 points within the cycle where sync can occur, so power phase will never be more than +- 30° out of sync, even before adjusting the alternators. So this means syncing 5000 or 10000 Km away is no different than 833 Km. For many years the LORAN C navigation network was used as a time base for the telecommunications industry to sync the telecom network. However, it only provided an accurate time base clock, not time of day.
On 2015-03-15 10:48 AM, James Knott wrote:
With 3 phase power, there are 6 points within the cycle where sync can occur, so power phase will never be more than +- 30° out of sync
I may have misspoken a bit about the purpose of the GPS. It's for keeping the phasor* measurement unit (PMU) in sync. These units report back to the system operator so that early signs of grid problems (and by early, that would be a few seconds) can be managed remotely. PMUs were required after the 2003 blackout: there was no shortage of diagnostic information after that outage, but most of it had imprecise time stamping, so was of little use to the postmortem investigation. Here's the PMU that every wind and solar project seems to use: https://www.selinc.com/synchrophasors/ It's not as pretty as the Synchronome clock that hangs in the board room of the IESO in /redacted/ (one is not supposed to know where their HQ is: it has no sign, is heavily armoured, runs on positive pressure, etc.) which was used to manage The Hydro's grid. The SEI unit is a little easier to interface to than the Synchronome's invar pendulum, though. Large solar inverters synthesize their power waveform by copying the incoming waveform. Unless they are tightly controlled, they'll happily act as little repeaters for any sync problems. cheers, Stewart *: Steinmetz for the win!
On 03/15/2015 04:58 PM, Stewart C. Russell wrote:
It's not as pretty as the Synchronome clock that hangs in the board room of the IESO in /redacted/ (one is not supposed to know where their HQ is: it has no sign, is heavily armoured, runs on positive pressure, etc.) which was used to manage The Hydro's grid.
I have been inside there a few times with my work. It's located at Lakeshore Rd. and Winston Churchill, in Mississauga. It's hardly a secret though. I knew about it years before I ever set foot in the place. You forgot to mention the bollards that are dropped down to allow vehicles to enter the property.
Stewart C. Russell wrote:
On 2015-03-14 10:09 PM, David Thornton wrote:
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete.
Accuracy numbers in PTP are measured in nanoseconds, while the NTP world talks milliseconds. So if you have that sort of accuracy requirement (power system phasors, etc) you probably want PTP with the appropriate dedicated hardware, while garden-variety Unix admins just want log records to show up with the right second. There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too. Meanwhile, ntpd is getting a bit long in the tooth[0] and the ntimed project looks interesting. [0] "support for hardware clocks EBay has never heard of" -- Anthony de Boer
"There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too." You can do PTP over IP but I understand you can also deploy "dedicated" networks that JUST do PTP ( no ip ). It is the dedicate nets you are talking about ? The sexiest PTP hardware I've seen is the solarflare stuff. David On Sun, Mar 15, 2015 at 5:58 PM, Anthony de Boer <adb@adb.ca> wrote:
Stewart C. Russell wrote:
On 2015-03-14 10:09 PM, David Thornton wrote:
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete.
Accuracy numbers in PTP are measured in nanoseconds, while the NTP world talks milliseconds. So if you have that sort of accuracy requirement (power system phasors, etc) you probably want PTP with the appropriate dedicated hardware, while garden-variety Unix admins just want log records to show up with the right second.
There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too.
Meanwhile, ntpd is getting a bit long in the tooth[0] and the ntimed project looks interesting.
[0] "support for hardware clocks EBay has never heard of"
-- Anthony de Boer --- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
PTP at IP (layer 3) still needs hardware-based (PHY or MAC level) timestamping in order to achieve O(nanoseconds) and O(ppb) performance in phase and frequency. The best performance is achieved with this support at every link that time sync data traverses. Telco nerds use Synchronous Ethernet besides... Mike On Mon, Mar 16, 2015 at 8:16 AM, David Thornton <northdot9@gmail.com> wrote:
"There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too."
You can do PTP over IP but I understand you can also deploy "dedicated" networks that JUST do PTP ( no ip ).
It is the dedicate nets you are talking about ?
The sexiest PTP hardware I've seen is the solarflare stuff.
David
On Sun, Mar 15, 2015 at 5:58 PM, Anthony de Boer <adb@adb.ca> wrote:
Stewart C. Russell wrote:
On 2015-03-14 10:09 PM, David Thornton wrote:
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete.
Accuracy numbers in PTP are measured in nanoseconds, while the NTP world talks milliseconds. So if you have that sort of accuracy requirement (power system phasors, etc) you probably want PTP with the appropriate dedicated hardware, while garden-variety Unix admins just want log records to show up with the right second.
There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too.
Meanwhile, ntpd is getting a bit long in the tooth[0] and the ntimed project looks interesting.
[0] "support for hardware clocks EBay has never heard of"
-- Anthony de Boer --- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
--- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 As an aside, I would love to see a presentation on keeping time in sync on Linuxen, with a deep dive into the innards of NTP. Sounds like GTALUG has the talent pool for such a presentation... - --Bob. On 16/03/15 12:33 PM, Mike wrote:
PTP at IP (layer 3) still needs hardware-based (PHY or MAC level) timestamping in order to achieve O(nanoseconds) and O(ppb) performance in phase and frequency. The best performance is achieved with this support at every link that time sync data traverses.
Telco nerds use Synchronous Ethernet besides...
Mike
On Mon, Mar 16, 2015 at 8:16 AM, David Thornton <northdot9@gmail.com> wrote:
"There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too."
You can do PTP over IP but I understand you can also deploy "dedicated" networks that JUST do PTP ( no ip ).
It is the dedicate nets you are talking about ?
The sexiest PTP hardware I've seen is the solarflare stuff.
David
On Sun, Mar 15, 2015 at 5:58 PM, Anthony de Boer <adb@adb.ca> wrote:
Stewart C. Russell wrote:
On 2015-03-14 10:09 PM, David Thornton wrote:
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete.
Accuracy numbers in PTP are measured in nanoseconds, while the NTP world talks milliseconds. So if you have that sort of accuracy requirement (power system phasors, etc) you probably want PTP with the appropriate dedicated hardware, while garden-variety Unix admins just want log records to show up with the right second.
There is PTP support in Linux, but you need an Ethernet card with the right timestamping feature plugged into a switch that does PTP too.
Meanwhile, ntpd is getting a bit long in the tooth[0] and the ntimed project looks interesting.
[0] "support for hardware clocks EBay has never heard of"
-- Anthony de Boer --- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
--- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
--- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.15 (GNU/Linux) Comment: Ensure confidentiality, authenticity, non-repudiability iEYEARECAAYFAlUM16kACgkQuRKJsNLM5eoKhgCg/zqAXoZpc5f/M5l1eo4SrZKr EXsAn0UERQBOidCsekZvI2hrYVkbn7bD =9YUp -----END PGP SIGNATURE-----
On Sun, Mar 15, 2015 at 10:36:10AM -0400, Stewart C. Russell wrote:
Absolutely. At the solar plant I audited the other week, I was amused to see that the great big huge expensive grid protection and control box was controlled via GPS sync. It's the cheapest way to maintain grid frequency at the far end of the grid. And I mean /far/; 5 hours west of Thunder Bay …
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so will likely be fiddly but complete.
IEEE 1588 I believe. No idea if there is an ISO standard for it. -- Len Sorensen
On March 24, 2015 4:24:31 PM EDT, Lennart Sorensen <lsorense@csclub.uwaterloo.ca> wrote:
On Sun, Mar 15, 2015 at 10:36:10AM -0400, Stewart C. Russell wrote:
Absolutely. At the solar plant I audited the other week, I was amused to see that the great big huge expensive grid protection and control box was controlled via GPS sync. It's the cheapest way to maintain grid frequency at the far end of the grid. And I mean /far/; 5 hours west of Thunder Bay
Sort of timing related; here's a neat desktop demonstration video of time honored lightbulb syncing of a polyphase AC grid. https://m.youtube.com/watch?v=RGPCIypib5Q
My finance client's use ptp not ntp.
Is PTP particularly hard to use? It looks like an ISO standard, so
will
likely be fiddly but complete.
IEEE 1588 I believe. No idea if there is an ISO standard for it.
-- Len Sorensen --- Talk Mailing List talk@gtalug.org http://gtalug.org/mailman/listinfo/talk
-- Sent via K-9 Mail.
participants (8)
-
Anthony de Boer -
Bob Jonkman -
David Thornton -
James Knott -
Lennart Sorensen -
Mike -
R. Russell Reiter -
Stewart C. Russell