NTP Reliability Proposal for Australia, Canada and New Zealand




Canada, Australia and New Zealand do not exactly have the most internal reliable national level NTP protocol networks. This inevitably has some technical repercussions that are not helpful to the general public.

Although it is possible to actively leech NTP time synchronizations from the limited number of universities and governmental institutions, the current number of freely available NTP servers too low in these nations to be in the national interest.

Leeching NTP time from the Asia-Pacific region has its limits in the Southern Hemisphere. Canada's ability to leech from US based NTP servers is helpful, but in the remoter parts of Canada this does not work well.

Clearly, the most of the world's local NTP networks need to be improved. However, this clearly must be done with as little cost as possible to government as reasonably possible. Since 2007 there has been an ongoing global economic downturn, and this downturn has limited the ability for any nation to arbitrarily build a dense high quality Stratum 1 time dissemination network.





The Post Office Solution


H
istorically, the Post Office in Australasia was a source of time signal dissemination from the 1890s to the 1970s.

Both Australia and NZ post offices exited the time dissemination business in the 1970s as the CSRIO and NZRO had essentially taken over the role. Very little has changed since the 1990s when GPS (and GNSS navigation technologies) took over the role of time signal dissemination.

An undersea (and land based) fibre optic cable
infrastructure suitable for the dissemination of high resolution time signals has only been in place since the early 2000s, and its coverage is not perfect. However, the overall dependency on exotic technologies that originate outside Australasia to disseminate time signals (and frequency standards) has been an ongoing problem for the past 50 years.

In Canada the Post Office was not significantly involved in time signal dissemination since confederation as the railroads and telegraphy companies provided the service. The NRC and its predecessor government agencies only became involved in national time signal dissimulation in the 1930s. In Canada, private Telecom entities providing time signals has been more of rule for the past 120 years, with the government providing the master clock signals only since  the 1940s.

In the current era, GPS (and GNSS) technologies coupled with fibre optic cable links are the way time signals are delivered in Canada -- but it is a patchwork system that ultimately is dependant on a similar nearby US time dissemination infrastructure.

Anyway you look at it, the Post Office -- be it a semi-Private entity or purely a Public owned utility -- is best suited to deliver publicly available time signals via a network of time dissimulation devices. The Post Office meets this criteria as it can disseminate time (of Stratum 1 and Stratum 2) at the lowest possible marginal cost with the largest area of redundant coverage.

There needs to be ongoing research with respect to creating local time delivery technologies (of Stratum 1 or Stratum 2 quality) that are separate from those technologies created outside Australia, NZ or or Canada. This research should be done with open sourced software as well as open sourced hardware. All the technology development should be accessible to the public.





Requirements


How many servers does each nation need, and where should they be?

Servers (and maximal permitted NTP data rate)
  • Australia : ~40 x 20 kbs
  • Canada : ~30 x 24 kbs
  • New Zealand : ~20 x 22 kbs
  • Ideally each nation should scale the number of post office NTP servers up to 50, but not exceed 50. There are myriad server management problems with having more than 50 servers.
  • Ideally if it is decided to implement 50+ NTP servers, the individual server data rate should drop to 24 kbs.


What server designs are optimal?
  • To reduce cost as much as possible they should be recycled Post Office computers, or Point of Sale (POS) terminals wherever applicable.
  • The NTP servers should have no Hard Disk Drive (HDD), but instead boot from a Flash Drive (FD). Most POS terminals are capable of this minor retrofit.
  • The FD boot and server drives should be managed and distributed by the Post Office internally.
  • The servers should either implement MINIX 3 or Linux, but with a minimal configuration (under 16 MB).
Are there any recommended practices that should be considered to increase the reliability of the the implementation?
  • The NTP logfiles should be uploaded 3 days, so as to help spot bad configurations.
    • The Post Office "DNS Zone File" (part of its DNS domain record) should be modified to account for this.
    • Each post office NTP server should also be mapped via DNS to the {*}.pool.ntp.org domain pools.
    • {*} implicitly implies [ntp##].post.[au | ca | nz].
      • The "post" is the post office Domain Name.
      • The symbol '##' implies 'forced zero' for formatting and sorting.



Infrastructural issues


What are the staffing requirements for this service?
  • Australia : One part time person per state, with NT and SA using the same person.
  • Canada : One part time per province, but with Yukon, NWT and Nunavut sharing the same IT person.
  • NZ : One part time person, with an adjutant academic (either on a stipend or actively involved in Computer Science research).
What modifications would be needed at the Post office websites?
  • Each Post Office website should provide information on accessing the constructed NTP network.
  • Each Post Office website should provide some real time monitoring data where applicable.


Technical References


NTP


Post Offices


Technologies needed


Related entities
  • Australia : TBA
  • Canada : TBA
  • NZ : http://ntp.net.nz/ (a Domain name entity that runs infrastructural DNS servers for Asia-Pacific use)



Created by
Initial document
Last revised
Last Revision

Revision State


Max Power
18 May 2010

12 September 2013

Minor text updates

Version 0.9b