[comp.protocols.time.ntp] General help-info

jitu@pivot.sbi.com (Jitu Davda) (06/02/91)

This is a standard posting requesting help/suggestions/guidance
from fellow netters using the NTP time deamons. I need to address
the following issues  ;

- I need to syncronize over 500 machine - almostly all
  exclusively Suns (490's, 330's and SparcStations at 4.1 & 4.1.1). 
  Since we also have International offices (London, Tokyo etc),
  I am looking for help on configuring such a large network and
  major pitfalls to avoid from people who may have already done
  large sites. ( help from people who have done smaller configurations
  will also be appreciated !) 
 
- I hear Suns are machines with 'bad' clocks. How bad are they
  really ? 

- Radio clocks, syncronization from internet stratum 1 hosts 
  or other external sources such as time observatories.  
  Should I be using one, two or a combination to provide
  good standard time keeping ? 

Again, Thanks for any help.

Jitu Davda                       | UUCP:  !pivot.sbi.com!jitu |Work(NY) 212-783-7708
Salomon Brothers, Pivot         
7 World Trade Center, 39th Floor
New York, NY  10004              | Internet: jitu@pivot.sbi.com 

edward@TWG.COM ("Edward C. Bennett") (06/03/91)

>- I need to syncronize over 500 machine - almostly all
>  exclusively Suns (490's, 330's and SparcStations at 4.1 & 4.1.1). 
>  Since we also have International offices (London, Tokyo etc),

This strikes me as interesting...Suppose a company has offices on
several continents and they want to synchronize their clocks. Do
they really need to peer with each other? Why shouldn't the European
office peer with a Stratum 1 server in Europe, Japanese office to
a server in Japan, etc.?

Look at it like this, the shortwave radio signals that the Stratum 1's
listen to is global, in effect an international network. Once the
signals are received and put on a wire it is redundant to send them
back around the world. Tokyo doesn't need to peer with New York on a
wire because WWV does it for them through the air.

-- 
Edward C. Bennett - The other WIN/MHS & MMDF guy	edward@twg.com
The Wollongong Group					(415) 962-7252
1129 San Antonio Road, Palo Alto, CA 94303
   "He's become a growling, snarling mass of white-hot canine terror"

mouse@LIGHTNING.MCRCIM.MCGILL.EDU (der Mouse) (06/03/91)

>>- I need to syncronize over 500 machine - almostly all exclusively
>>  Suns (490's, 330's and SparcStations at 4.1 & 4.1.1).  Since we
>>  also have International offices (London, Tokyo etc), [...]

> This strikes me as interesting...Suppose a company has offices on
> several continents and they want to synchronize their clocks.  Do
> they really need to peer with each other?  Why shouldn't the European
> office peer with a Stratum 1 server in Europe, Japanese office to a
> server in Japan, etc.?

It depends on what's most important to them: that their clocks be in
sync with one another or that their clocks be in sync with The Real
Time?  If the former, they should peer with one another, with the top
internal stratum peering with something outside; if the former, they
should do as you suggest.

Of course, your suggestion also reduces international packet traffic;
this may or may not make any difference, depending on the situation.

					der Mouse

			old: mcgill-vision!mouse
			new: mouse@larry.mcrcim.mcgill.edu

Mills@udel.edu (06/03/91)

Edward,

You really do need to peer with several (like three) radio clocks,
because they sometimes lose their chime due to various ionspheric
phenomena and carbon-based unit failures. Also, the radio transmissions
used by various clocks are not available everywhere in the world and,
in fact for some services, not outside continental US and continental
Europe. In principle, GPS is available everywhere, but at $10K a pop,
not affordable by most.

At the moment there are stratum-1 servers in US, Canada, Switzerland,
Germany and Norway. Even Norway is a special case, since only its
timebase oscillator is synchronized to national standards and the
seconds are numbered using NTP and other stratum-1 servers. It would of
course be wonderful if heavy hitters with lots of hosts and bucks
could buy GPS receivers for timecode-deprived parts of the world and
make their ticks accessable via NTP.

Dave