Like with Google and Microsoft, it's very likely Amazon are already working on a similar platform. While they may not have an as-widespread datacentre (DC) distribution as Google (6500+ DCs worldwide of various sizes according to today's presentation) Amazon's overall capacity and distribution could be similar (fewer, larger DCs). They have AmazonStore, Prime, Twitch, Audible, etc already using those DCs.
FDev could have the option of porting Elite Dangerous to the AWS Cloud Product (AWStadia?) and at the same time update current clients (PC, Xbox, Playstation) with hybrid tech from that porting:
- Move the client-side peer-to-peer networking to Cloud Instances/Nodes (gonna call them nodes from now on, to avoid confusion with in-game instances, i.e. ~32 in bubble, 100+ during mass-jumps)
- Windows/Xbox/Playstation clients install/play the game as normal
- Clients connect and communicate with their nearest node - small latency between client & node
- Nodes use Peer2peer to connect to other AWStadia nodes - minimal latency between nodes (either on same node, or ones nearby)
- All nodes connect to the main Elite Dangerous AWS servers (BGS, authentication, adjudication, missions, etc) - minimal latency between node & server
- Clients still do the vast bulk of performance-requiring tasks - rendering, input-recognition, gameplay calculations, etc
- Bar the connectivity of client to Cloud Instance/Node, there shouldn't be any Peer2Peer issues like we have now (incompatible ISPs & MTUs, UPNP, port tunneling, etc) as they'd all be "server-side"
- Server-nodes/instances distributed worldwide means much lower networking latency than if all players connected to one/few separate worldwide servers (e.g. EU, USA, Oceania), so it'd be close to the low-latency-benefits of current Peer2Peer
- Lower bandwidth requirements of clients, as they'd only be sending/receiving data between themselves and the nearest node, instead of to/from all players in the same game instance
- Performance and Bandwidth requirements of Elite Dangerous nodes would be dramatically lower than Stadia-equivalent nodes - only running the Elite Dangerous networking stack
- Therefore, each node could probably support thousands of connected-clients
- VR continues to be supported on clients
- Possibility of game improvements that require server-side networking: larger more-reliable instances (100s of players?) and matchmaking, persistent NPCs, immediate server-wide events & effects, combat logging punishment, etc and all the emergent gameplay that would come from those changes
Hierarchy:
TL;DR: Amazon have all these datacentres worldwide, a networking rewrite and upgrade would be very beneficial, and Elite's next major expansion (it's biggest one yet) is in development for 2nd half of 2020, so the time is right to do it