[Starlink] Jamaican Starlink Outages and a hint of shared infrastructure

[Ulrich Speidel]

On 25/06/2024 3:42 pm, David Lang wrote:
> Ulrich Speidel wrote:
>
>> This can also lead to weird effects globally. For example, much of 
>> the traffic between Japan and New Zealand *could* in principle 
>> trundle down to Guam and from there to Sydney and then to Auckland. 
>> Which would be kind of shortest path. And occasionally it does. But 
>> just as often, you see it crossing the Pacific to the US West Coast 
>> (or from Guam to Hawaii) and from there back to New Zealand. Why? 
>> Good question. Was it because US backhaul carriers were cheaper for a 
>> while with the US dollar being soft and the Australian / NZ 
>> currencies surging in comparison? Were there government incentives 
>> for carriers to let traffic run through US territory for intelligence 
>> access (if so, the NSA would have to fear
>
> It's important to realize that BGP doesn't know how long any link is. 
> it defines 'closest' by the number of hops.
>
> so Japan -> LA -> Auckland is 'shorter' than Japan -> Guam -> Sydnes 
> -> Auckland even though it's much longer, probably through more 
> congested links, and higher latency.

Indeed, but that's assuming that (a) there are hops (just because two 
cables land on the same island, it doesn't mean that they are connected 
there) and (b) that pure BGP hop count metric isn't overruled by the 
sending ISPs or someone along the way, as it often is. E.g., Hawaii to 
NZ is just one cable "hop", but traffic from the University of Hawaii to 
Auckland also used to go via the US West Coast for a long time 
(someone's fixed it now) as they didn't peer with Southern Cross.

-- 

****************************************************************
Dr. Ulrich Speidel

School of Computer Science

Room 303S.594 (City Campus)

The University of Auckland
u.speidel at auckland.ac.nz
http://www.cs.auckland.ac.nz/~ulrich/
****************************************************************