The problem is that i don't know too much about these new sat networks. They are supposed to offer coverage for undeveloped areas. Connecting UE xTRs, to proper internet router xTR. Maybe the operator wants to keep them as simple as possible or as IP relays.
So potentially yes, no lookups, just algorithmic mapping mac to tile, and tile to IP. If this algorithm takes a micro sec, it should be good for a few seconds - tile size vs sat speed. The problem of source routing this “GTP” tunnel “UE to UPF” xTRs is the number of headers. Perhaps it should be established every few seconds by the UPF xTR and labeled in the route. --szb Cell: +972.53.2470068 WhatsApp: +1.650.492.0794 > On Apr 5, 2022, at 00:01, Dino Farinacci <[email protected]> wrote: > > >> >> Ok fixed. > > Thanks, looks good. Cached for reference. > >> Tiles have dimensions so better then GPS, but if these 1000s of new sats >> move in completely different general tiers around the earth, we can >> enumerate them and add to the H3IP. Theres spare bits cause smaller the city >> size tiles are irrelevant for this app - unlike the vehicle lisp app. In any >> case H3ID is just 64bits. > > Okay. Did you want to algorithically imply the IP address from the H3ID and > vice versa? That could avoid another lookup. And requires IPv6, but that is > okay. > > So another issue. The IP address will change as the sat moves. That's okay > becuase those H3IP addresses are not RLOCs and don't go in the mapping > system. So we don't have sat mobility entropy. > > But now comes the questions is why does an xTR need to know the IP address of > a satellite node? > > A GS-xTR can just send on RF-link and whatever satellite up there that > receives signal can forward packet. If there are multiple satellites > receiving the packet, we only need one to forward. This could look like a > stub LAN where VRRP decides who is default router allowing multiple GS-xTRs > to use the same satellite if they are in the same field of view. > >> I don't know how the RF works, but if we know at each given moment which >> sat-mac is in which H3IP we can have the GSxTR source plot the segments. >> Overlay can be a good way to drive a dynamic underlay, connecting the ground >> IPs through the sky IPs. Anyway interesting domain if its not already >> solved. > > I think if you do the plot at head-end, you aren't going to get reachability > if the path changes in packet mid-flight. You can get away with > source-routing in an ISP network because the topology doesn't change that > much and there aren't as many path options like there will be with satellite > networks. > > So I think we should just send an encapsuluate packet up and let the > satellite network routing (many approaches exist and are being used) do its > job. > > Dino > > _______________________________________________ > lisp mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/lisp _______________________________________________ lisp mailing list [email protected] https://www.ietf.org/mailman/listinfo/lisp
