[time-nuts] Low-long-term-drift clock for board level integration?

Hal Murray hmurray at megapathdsl.net
Tue Feb 21 01:24:15 UTC 2012


brooke at pacific.net said:
> It was my understanding that the receiving station knows the path taken, is
> that the case? Or are you saying even when  it's the same path the time
> delay has large variations? 

Even if the path is stable, the delays vary due to queuing delays in routers.

The simple example is a router with 3 cables.  Call them A, B, and C.  Assume 
they are all the same speed.

If B is idle, packets that come in on A can go out C right away.

But suppose A and B are both sending traffic out C.  If packets arrive at 
close to the same time, the second one will have to wait on a queue until the 
first one is finished.  If a clump of packets arrives the queue can get 
longer.

There is another source of queuing.  That's when the link speed changes.  
Suppose the server has a 100 megabit connection, the backbone has gigabit 
links, and the last hop is 1 megabit.  (just to pick some round numbers)  If 
the server sends a clump of packets, they go out at 100 megabits, probably 
won't hit much queuing delay in the backbone, but then they pile up on the 
last hop.



-- 
These are my opinions, not necessarily my employer's.  I hate spam.






More information about the time-nuts mailing list