# Physically induced latency in internet connections

As a trade I am a software architect. It does happen that I have to design sites where the servers are located in different parts of the world.

Now an essential parameter to consider is latency: that is the time it takes between when a packet is sent from one server, and when it is received on the other side.

Of course, the packet might go through other systems etcetera. However, the hallmark of a good connection is low latency.

Question: what is the minimal latency that can be obtained over a coaxial wire (i.e. what is the maximum speed reachable by an electrical impulse over a coaxial wire)?

Question: what is the minimal latency that can be obtained over a fiber optic connection (if different from distance/c)?

-
Your title mentions relativity, but your question body does not. Could you resolve this for us please? – Noldorin Jan 5 '11 at 20:31
I assume that the maximum speed of a signal through a medium is c, hence a relativistic limit - do change if you think it's unappropriate. – Sklivvz Jan 5 '11 at 20:44
I can say that the latency with a fiber optic connection would be over distance/c because of all the processes the incoming signals will go through in the receiver and the outgoing signals will go through in the transmitter. It would be just a few microseconds of difference from distance/c though. Besides light does not follow a direct route inside the cable itself. – Cem Jan 5 '11 at 21:46
@Cem, I am looking for a lower limit in order to compare performance regardless of the distance. – Sklivvz Jan 5 '11 at 22:12
@Sklivvz: Ah, I see. Yeah, the speed of light is certainly involved, but there's no relativity really. I understand your question now anyway... – Noldorin Jan 5 '11 at 22:22

http://en.wikipedia.org/wiki/Wave_propagation_speed

79% for coax, 67% of speed of light for fiber. So it's 42ms for 10'000km of coax, and 49.7ms for 10'000km of fiber.

For roundtrip time multiply by 2.

Reality adds some more latency (up to 1.5-2x) on transcievers, queues on overloaded segments and non-optimal paths.

Update: As a side note, I really doubt someone could make 10'000km link at 10Gbit+ via coax :-D Signal amplification is just much easier & cheaper in case of fiber.

Update2: There is also exotic fiber with empty core, which have almost 100% speed. I've heard this would be used to pass trading information across continents to gain additional profit. 10ms worth alot of money :-D

-
I couldn't find the 67% percent figure in the linked article. Are you sure that fiber is slower than coax? – Sklivvz Jan 6 '11 at 15:53
For fiber that was different article. Speed of light in fiber is c/refraction of fiber core, which is ~1.48. 1/1.48 = 67% – BarsMonster Jan 6 '11 at 16:44