Multiplexing chain synchronization across multiple wallets

[abailly]

• Synchronizing 100s of wallets introduces a lot of contention on resources, particularly IO and CPU, because each wallet manages their own connection to the node, deserialised blocks, etc
• Most of the time, those wallets are following the chain at the same or very close point in time, so they would just need to share the result of RollForward/RollBackward messages if they were able to share the same connection
• This leads to the simple idea of Chain-Sync Multiplexer whereby multiple clients can share a single connection:
  • each client has their own "cursor" (eg. Point) they track
  • each connection has also their own cursor
  • multiple clients interested in the same (or close) point just can share the same connection
• this requires some component that makes the wallet believe it's talking to a node, so implementing ChainSyncServer while it maps the clients request.

[HeinrichApfelmus]

Most of the time, those wallets are following the chain at the same or very close point in time, so they would just need to share the result of RollForward/RollBackward messages if they were able to share the same connection

Nice. Just to spell it out, the intended gain in efficiency would come from being able to share an in-memory pointer to a data structure representing a block as opposed to deserializing multiple in-memory copies of the same block.

[abailly]

that's my understanding so far yes. With the current system, it seems this sharing is not possible: every single instance of wallet has to own their own connection, do the deserialisation, before being able to scan the txs for addresses they care about.