Transatlantic response times.

I wasn't really sure where to post this, but I figured NANOG would have some
insight or at least experience here.

I was curious if anybody would share what they consider to be average or
acceptable transatlantic ping response times over a T1.
I know there are tons of variables here, but I am looking for ballpark
figures.
Assume that utilization on the circuit is extremely low, and you are
measuring point to point across the line. You can also assume no other
bottlenecks effecting the response times (router performance, or what not).
Should you see a ~150ms trip? 250ms? 450ms???

Also, if possible, include the to and from info. Obviously NYC to London
is a bit different than Dallas to Prague or something.

Is there any equation to estimate response times? For example, if your
circuit from A to Z has a 500ms avg response, than that equates to a circuit
distance of aprox. 5000 miles or something?

Thanks in advance,

Mike

London to NYC 70ms RTT

London to Seattle 150ms RTT

I was curious if anybody would share what they consider to be average or
acceptable transatlantic ping response times over a T1.
I know there are tons of variables here, but I am looking for ballpark
figures.
Assume that utilization on the circuit is extremely low, and you are
measuring point to point across the line. You can also assume no other
bottlenecks effecting the response times (router performance, or what not).
Should you see a ~150ms trip? 250ms? 450ms???

Something like 70 - 100 ms with small packets.

Is there any equation to estimate response times? For example, if your
circuit from A to Z has a 500ms avg response, than that equates to a circuit
distance of aprox. 5000 miles or something?

The three main components in the delay are:

- serialization delay: it takes a certain amount of time to get a packet
  out of the interface. This is the size of the packet divided by the
  bandwidth of link. For instance: 1500 bytes = 12000 bits / 1536000 bps
  ~= 8 ms. (Double for RTT.)

- speed of light: this depends on the medium. For fiber, it's about
  200,000 km/s = 125,000 mi/s. So 5000 miles worth of fiber (which could
  be the atlantic, but your milage may vary) is 40 ms. (Double for RTT.)

- queuing delays: this depends on how busy the circuit is and on the
  number of hops.

You can remove the queuing factor by leaving your ping running for a
fairly long time and then only look at the shortest RTT. If the shortest
and the average RTTs are far apart, the circuit is very busy.

As I'm sure you remember from your physics class, light travels through a
vacuum at 299,792 km/sec (lets call it 300,000). When it travels through
other mediums, it moves slower based on that medium's refractive index.
For example, water has a refractive index of 1.33, which means light
travels through water at about 0.75c, or about 225,000 km/sec.

Fiber works on a principal called "total internal refraction", which means
that the light is continually reflected into the core with no (or little)
loss in the cladding. To accomplish this, different material with
different refractive indexes is used. Since the cladding has a lower
refractive index than the core, as long as the angle of incidence exceeds
a critical angle, the light will be reflected back into the core instead
of shooting out the sides. The values of refractive indexes used in fiber
are usually something along the lines of 1.46 in the cladding and 1.48 in
the core.

So if you do a little math, you'll see that light propagates through fiber
at around 0.67c, or 200,000 km/sec (or approx 125,000 miles/sec). Putting
that in ms terms so even ping monkeys understand, you get approximately
1ms of speed-of-light induced delay per 200 km (or 125 miles) of fiber
path.

As an example (and to answer your original question), 500ms RTT / 2 is
approximately a 50,000 km or 31,250 mile fiber path. Adjust a little for
all the microseconds of switching and buffering which happens to your
packet along the way, and you can get a fairly good idea how drunk the
people were when they laid your fiber.