I would prefer a setup where there's in effect a size cap on core service provision nodes; one per million people.
Nodes are competitively ranked for reliability and value delivered (as a means of controlling the pay and promotions of the people in charge of them) and more or less forbidden from using the same hardware or software as their failover nodes. (more or less = unrelated linux distros probably OK, linux versus BSD definitely OK, all versions of IBM mainframes are the same thing, etc. It would take some taxonomic effort.)
Every failover connection uses different cable runs, in the sense of being physically separate. I don't think we could argue for keeping some copper around but presumably there's alternatives for types of fibre depending on range.
All the nodes have to present the same access protocols, though; the layer of actual ISPs, the people you or I would contract with to get internet, all use the same interfaces on both sides. Hopefully this makes switching easy; hopefully it allows ISPs to have buddy agreements where is the ISP's hardware fails they can fail over to an adjacent ISPs and the customers don't notice.
Northern regions might be a special case, but the likely price point for long-duration solar high-altitude loitering aircraft looks low enough to fly redundant backups. (And to have some surge capacity on hand in adjacent regions; those aircraft aren't fast but have approximately infinite range once they're up there.)
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
timeasmymeasure
no subject
Date: 2022-07-09 06:29 pm (UTC)I would prefer a setup where there's in effect a size cap on core service provision nodes; one per million people.
Nodes are competitively ranked for reliability and value delivered (as a means of controlling the pay and promotions of the people in charge of them) and more or less forbidden from using the same hardware or software as their failover nodes. (more or less = unrelated linux distros probably OK, linux versus BSD definitely OK, all versions of IBM mainframes are the same thing, etc. It would take some taxonomic effort.)
Every failover connection uses different cable runs, in the sense of being physically separate. I don't think we could argue for keeping some copper around but presumably there's alternatives for types of fibre depending on range.
All the nodes have to present the same access protocols, though; the layer of actual ISPs, the people you or I would contract with to get internet, all use the same interfaces on both sides. Hopefully this makes switching easy; hopefully it allows ISPs to have buddy agreements where is the ISP's hardware fails they can fail over to an adjacent ISPs and the customers don't notice.
Northern regions might be a special case, but the likely price point for long-duration solar high-altitude loitering aircraft looks low enough to fly redundant backups. (And to have some surge capacity on hand in adjacent regions; those aircraft aren't fast but have approximately infinite range once they're up there.)