[time-nuts] can of worms: time-of-day in a community radio station

[Hal Murray]

> My understanding is that it would work moderately well even without a
> stratum-1 server, at least be able to operate within a few tens of
> milliseconds for several hours. Although I confess I haven't used peering in
> a very long time. In my workplace we added stratum-1 GPS symmetricom NTP
> servers about 6 months after the above mentioned setup.

Suppose you start with 2 stratum 2 servers connected to different stratum 1 
servers on the internet and also to each other.

Suppose one of the stratum 1 servers goes down.  After 8 poll intervals with 
no response, your server using it will switch to using your other local 
server.  That puts it at stratum 3.

Suppose it was the link that went down rather than the server.  (It was a race 
as to which of your servers timed out first.)  Soon your second server gives 
up and switches to using your first server.  The first server is now stratum 
3, so the second server becomes stratum 4.  Then the first server becomes 
stratum 5 and the second becomes 6 ...  After a few more iterations, one gets 
to stratum 16 which is a no-data.  The other soon times out.

If you have evidence that something better happens, please clue me in.

You can keep going with either a localclock or orphan mode.  If you do that, 
you are running what GPSDOs call holdover.  It works well enough for a while.  
How long is a lot of it-depends.


> My apologies, I just did a quick googling because I was on the phone and
> didn't have access to my desktop bookmarks. This is the list I normally use

> http://support.ntp.org/bin/view/Servers/StratumOneTimeServers 

That list is better than most.  I would still treat it with a  reasonable 
amount of caution.  Scan the column of last-modified dates.

Don't get hung up on stratum 1.  A stratum 2 right next to a stratum 1 can 
keep time well enough so that you will have a hard time telling the difference 
over a network connection.

Cloudflare offers  non-smearing public servers.
  https://blog.cloudflare.com/secure-time/

----------

For "interesting" reading, check out the Wikipedia page on NTP abuse.
  NTP server misuse and abuse
  https://en.wikipedia.org/wiki/NTP_server_misuse_and_abuse

I think Dave Plonka's writeup of the U Wisc event should be required reading 
for any computer science program.  (It's in the References section of the wiki 
page.)




-- 
These are my opinions.  I hate spam.