Re: [manet] Routing in emerging LEO satellite networks

["[e-mail]" <[e-mail]>]

Hi Henning:

Thanks for the reply. It a good question : )

In an idea world, satellites should work in their orbits in plan and have predictable trajectory (as well as the connectivity). 

But something might change in the merging LEO satellite networks.

First, satellite movements are somehow "predictable", but are not so "deterministic" as we may expect.

The outer space is a very complex environments. Due to the lower altitude, satellites are affected by the gravity, and satellites have to continuously adjust their orbit altitude and position during the orbiting time. Just like self-driving cars, satellites have to monitor debris nearby and accordingly locally adjust their orbits to avoid unexpected conficts. 

In addition, there are a range of unexpected factors such as radio interference, solar storm, channel fading, bad weather conditions, etc, that could change the inter-connectivity and link quality for inter-satellite and ground-satellite communications.

Second, even though we exclude these "unexpected" factors, it is still very chanllenging to pre-calculate all possible any-to-any routes on a global scale.

Note that emerging LEO constellations like Starlink and Kuiper have thousands of LEO satellites. And the entire network topology changes frequently and endlessly (sometimes it changes in tens of seconds!). If we try to pre-calculate all routes for all topology snapshots, there should be a significantly large number of possible routes, for every any-to-any communication pair, and in every time slot! Pre-caching all possible routing tables on each satellite could not be doable for resource-constrained satellites.

So the new things here are globe-wide LEO dynamics (predictable and unexpected), 3D network topology (when integrating LEO satellites into terrestrial Internet), and significantly increase in the constellation sizes.

That is why we say that it is challenging to make a scalable and robust  routing for the emerging LSNs. And we believe there should be many new problems and questions : )

Best regards,
Zeqi Lai


[e-mail]
 
From: Henning Rogge
Date: 2023-11-08 02:59
To: [e-mail]
CC: [e-mail]; zhangqi; wuqian; lihewu
Subject: Re: [manet] Routing in emerging LEO satellite networks
On Wed, Nov 8, 2023 at 8:48 AM [e-mail]
<[e-mail]> wrote:
>
> Hi all MANET:
>
> Thanks for the nice talks in IETF 118. As a new attendee, I learnt a lot.
>
> I am currently an assistant professor at Tsinghua University, and our group are working on the internetworking technologies in the upcoming LEO satellite networks (LSN).
>
> One important direction of our research is exploring how to achieve scalable and robust routing in LSNs. And we have many recent efforts in this direction. Some have been published in the research conference of computer networks.
>
> Here are some of our recent research works:
>
> (1) Routing in LSN: https://ieeexplore.ieee.org/abstract/document/9651919
> (2) Space-Groud topology design: https://ieeexplore.ieee.org/abstract/document/9796886
> (3) Simulation platform for LSN (with open source implementation): https://www.usenix.org/conference/nsdi23/presentation/lai-zeqi
>
> We believe LSN has many similar features with conventional mobile ad-hoc networks, just like the high mobility, frequent handovers, unstable communication links, etc.
 
I am not sure about the similarities to adhoc networks... all orbits
of satellites in a LEO mega constellation are known, so each node can
calculate in advance when a communication link becomes available and
when it will fail.
 
This means you can calculate a path towards the target without ever
doing something like neighbor or path discovery, just by knowing the
time and all relevant orbits. Or maybe I am missing something?
 
Henning Rogge