Internet dies if GPS dies?

[Tyler Durden]

Michael Shields wrote...

"I'm not a SONET guru, but my understanding is that even with the
internal clocks, you would see at worst an occasional error burst for
a few ms.  No credible engineer would make a network that fell over
without GPS anyway, since it's just too easy for someone to
accidentally knock over your antenna while installing another, or nick
the cable with a saw, or who knows what.  In a nutshell, this isn't
something I'd worry about."

SONET Synchronization is not a thing that can easily be discussed in any 
short amount of time. Synch in a SONET network is "provisioned", NE-by-NE, 
and if its done right then the chances of timing slips are very low.

However, in certain cases you can have a big timing "island", and in this 
case over the course of days you'll start to see more and more slips until 
that whole Island pretty much disconnects from the rest of the network. 
(This can happen when the primary reference fails, and when the reamining 
NEs all get their timing from each other in a big loop, like a snake eating 
its tail. The whole clock for this island begins to drift wrt the rest of 
the network, and in some cases there can eventually be enough slips as to 
cause some NEs to declare AIS. However, if Synch was provisioned correctly 
this won't be seen.)

The existence of possible timing islands, owever, should not be strongly 
associated with the loss of GPS. Traditionally, SONET networks use Cesium 
clocks as their primary reference source.GPSs have only started to 
proliferate in order to simplify synch. So in order for GPS cancellation to 
do anything at all to the internet, you'd have to have a network that has no 
access to ANY stratum1 clock (Cesium or GPS), you'd have to have old NEs in 
that network (ie, with a poor internal SONET clock), AND you'd have to have 
a timing island. And even then, only this island would be affected. If this 
island were in the middle of UUNet, you might see some slowdown.

As for routers I'll take it that they can read timing out of the Synch 
Status Byte (S1 in the SONET overhead), so they'll "know" what the problem 
is at least (ie, that byte will tell them what the quality of the synch is). 
If the line clock is bad, I'd guess that the big Cisco routers will go into 
"holdover" (or whatever router heads call holdover), and that's the time 
they'll use to stamp the flows with.

So, no problem I'm 99.9% sure. (And I'm rarely that sure of anything.)
-TD

_________________________________________________________________
Add photos to your e-mail with MSN 8. Get 2 months FREE*.  
http://join.msn.com/?page=features/featuredemail