Dear Nanog:
I was wondering if there is a benchmark for long-haul circuit latency... For example if I had a T1 circuit with 2900 miles between the two end-points (and assuming the provider is best case scenario) can I do something like (miles*latencyfactor) = 5 ms for 2900 miles?
Thanks,
Christopher
Hi Chris,
On long haul and reasonable bandwidth, I figure as a rough estimate to
divide the distance in kilometers by 100 will be a very close approximation
to an RTT in ms. For instance I get 147 ms ICMP RTT on the Southern Cross
Sydney-Seattle link where the fibre path is about 14700 km. It assumes a
transmission speed close to .66 the speed of light. The calculation using
miles is a little messier 
Bruce
"Christopher Wolff" <chris@bblabs.com> writes:
Dear Nanog:
I was wondering if there is a benchmark for long-haul circuit
latency... For example if I had a T1 circuit with 2900 miles
between the two end-points (and assuming the provider is best case
scenario) can I do something like (miles*latencyfactor) = 5 ms for
2900 miles?
it's gonna be a lot more than 5ms for 2900 miles. muxes and bit
regens introduce a tiny amount of delay. there's delay for clocking
the packet onto and off of the circuit (which becomes less significant
as the speed of the circuit goes up, but is still a couple of
milliseconds each way on a t1). remember that airline miles and road
miles have nothing to do with circuit miles, and it is not uncommon to
see a circuit go from dc to san francisco via new york, chicago,
dallas, and los angeles.
anyway, on a 2900 mile circuit the big delay factor is the speed of
light. in glass, this is 200 million kilometers per second, plus or
minus. that means you will never ever ever see less than 23
milliseconds each way. 46 for a ping.
i would say that for a t1 circuit between two places 2900 miles apart
anything under 70 ms is just dandy.
---rob
rs@seastrom.com (Robert E. Seastrom) writes:
light. in glass, this is 200 million kilometers per second, plus or
200 million meters (not kilometers) per second. But what's a few
orders of magnitude between friends?
---rob
Christopher Wolff wrote:
I was wondering if there is a benchmark for long-haul circuit
latency... For example if I had a T1 circuit with 2900 miles
between the two end-points (and assuming the provider is best case
scenario) can I do something like (miles*latencyfactor) = 5 ms for
2900 miles?
Speed of light is apprxoximately 2.998e+8 m/s (or 299,800,000 m/s),
which is approximately 186,287 miles/sec.
2900 mi / 186,287 mi/s = 0.01557s = 15.57ms. This is the absolute
minimum latency you can possibly get over that distance. Switching
performance and congestion will make your actual latency somewhat
larger.
-- David