- Aug 4, 2005
Congrats. You have caught up to something at least 3 of us posted already weeks ago in this thread.Not going to quote it all because that's one wall of text. My whole argument for the lower latency is based upon theoretical numbers and calculations as to what is possible given that it does the following for long backhaul transit.
User -> 380km Low Orbit satellite -> Backhaul Sattelite at 500-1100km -> around 5 hops over the transit satellites in the higher orbit. Given that the distance between Johannesburg and London is 9070km in a straight line distance. I will work out the actual orbital distance at 1000km when I have some time. So as an example im going to throw on 50% leeway so 9070*1.50 = 13605km distance.
Using the current speed of fibre it would cover that 13605 km in roughly ~138ms and that is with a 50% leeway on the distance, which might be somewhat more accurate than my previous calculations. Then if you going up and down 1000km each way that's 2000km which adds another 22ms. So total latency is 160ms from JHB to London. This includes a transit latency of 1ms per a satellite over around 5-6 hops.
I get that this is basically on par with current latency. I mean, if we managed to get a fibre path through Africa its 12,877km From Johannesburg via current African roads. Having a fibre backhaul here would give us 137ms (JHB -> London) and 117ms to Spain.
This is all assuming I am being a realist on how the tech will function and how there is a 99% chance it's useless for a low latency project like I am suggesting. But, hear me out. There is a chance.
Best Case Scenario for Starlink
Starlink as a backhaul provider from JHB Datacentre to London Datacentre.
I need to point out that Starlink is rolling out lasers as it will act as a major backhaul on inter-satellite connection, which lowers latency and the lasers actually are faster than fibre, don't forget that. Speed of light in space/vacuum. Light travels at 300,000 km/s in a vacuum, but when in glass it around 200,000 km/s. This means that it is technically 50% faster than what fibre would be. For the sake of SCIENCE, I would say we can assume it will be ~40% faster than current fibre optics in transition speed.
Johannesburg -> 380km Low Orbit satellite -> Backhaul Sattelite at 1100km -> around 8 hops over the transit satellites in the higher orbit. Given that the distance between Johannesburg and London is 9070km in a straight line. It would be around 13,000km given the extra distance of the satellites all not being in a line and the distance from the ground to orbit.
The latency on fibre over that distance would be 132ms with 1ms equipment latency.
On a laser backhaul (the 10,000km distance) it would be around 88ms + 22ms for the transit from the ground up, giving us a total latency from JHB - LONDON at around 110ms. Then your usual 2-5ms hop for the people in JHB and you can get a latency of 112-115ms.
Now another exciting point is the whole, a bunch of servers are located lower in Europe, such as Paris. Which is around 8728km at 20% leeway that's around 10473km = 106ms with 200 KM/millisecond fibre or 72ms with 299 KM/millisecond laser transit.
This is where I get excited.
Disclaimer: I am a crazy low latency person. But for interest sake, watch this video with a grain of salt.
The facts are really simple.
There will never be enough ISL capacity to make any of this a reality in the Starlink network for more than absolutely mission-critical traffic.
There is a far more likely scenario that ISL links between GSO satellites just might happen before LEOs can get there, and then there are only 3 satellites required to get around the globe.
Getting excited about the realities of the physics behind the speed of light is really, really old news like when Maxwell developed just about everything we know about electromagnetic waves.
If you want to see a total analysis of latency and how t can be used, read some of Isaac Asimov's books on truly long haul communications within our own planetary system.