[ntp:questions] Leap second bug?

David Woolley david at ex.djwhome.demon.co.uk.invalid
Thu Jan 3 22:57:14 UTC 2008

In article <9yVej.43847$UZ4.31290 at edtnps89>,
Unruh <unruh-spam at physics.ubc.ca> wrote:

> and the time required to timestamp that packet is about another 2usec, with
> fluctuations depending on whether other interrupts are being serviced.

On both Linux and Windows, interrupt latencies of more than 4ms and
even more than 10ms are quite common.  The interrupt processing time
from the idle loop is not a good indication of the timing on a system
doing real work.  (The above figures are based on clock interrupts 
overrunning at 250 Hz and 100 Hz clock frequencies, typically when doing
IDE disk I/O.)

> Actually if I believe tcpdump timestamping compared with the packet
> timestamping, it can at times be a few msec to get the packet out onto the
> net. Ie, the network card is far worse than the network itself. 

Network contention delays, which these almost certainly are, are normally
considered part of the network delay, not part of the network card delay.

> Most of my machines pass through two or three switches along the way ( all
> GBit switches by now)

I think most switches these days are store and forward, so will incur
contention delays at each stage, unless the network is seriously over-

> Fortunately Canada is very one dimensional. So stuff tends to go the same
> way there and back. Between Sask and BC the only alternatives are Calgary
> or Edmonton, and I suspect that the backbone goes through calgary.

Most asymmetric delay problems are due to network contention on the link
to the ISP, because many internet users are net consumers and tend
to have peaks at certain times of day when people tend to do their net
accesses.  That tends to result in severe asymmetry for a few tens of 
minutes, with the excess delay being in the downlink direction.  Again,
this is because delays on a properly dimensioned network are predominantly
due to contention, rather than serialisation or speed of light factors.

More information about the questions mailing list