PubSub Booster network integration

[vyzo]

For context, see protocol/web3-dev-team#104

Code related Action Items:

• Create bootster pubsub profile.
• Create nil coldstore abstraction for splitstore.
• Ensure splitstore works with a nil coldstore; complete or cherrypick Splitstore optimizations #5808 .

Deployment:

• Coordinate with infra to get machines with DNS and deploy code.
• Reach out to miners in China and establish direct peering agreements.
• Reach out to exchanges and establish peering agreements.

[f8-ptrk]

Reach out to miners in China and establish direct peering agreements.
Reach out to exchanges and establish peering agreements.

@vyzo can you elaborate a bit how these agreements would look like and what exactly they are aiming to "peer"?

[vyzo]

@f8-ptrk sure.

We want to peer the miner's full node (one of them suffices if they have multiple, they can do their own internal peering for their full nodes) with one of the boosters, and maintain this peering agreement indefinitely.

Basically we need to get their full node multiaddr and have them configure a respective pubsub booster for direct peering.

So let's say that we have full node /ip4/.../p2p/QmA peering with booster /dns4/.../p2p/QmBooster.
Then the full node will have to add this to its configuration:

[Pubsub]
  DirectPeers = ["/dns4/.../p2p/QmBooster"]

and the booster will have to add this to its config:

[Pubsub]
  DirectPeers = ["/ip4/.../p2p/QmA"]

that's all it takes code wise.

[NSC-FIL]

If this helps with increasing successful block rewards it is a very welcome change. Seems like lucky value for small and medium miners has dropped since Hyperdrive upgrade.

[jennijuju]

Seems like lucky value for small and medium miners has dropped since Hyperdrive upgrade.

Could you please provide the data source that drives up to this statement?

[f8-ptrk]

If this helps with increasing successful block rewards it is a very welcome change. Seems like lucky value for small and medium miners has dropped since Hyperdrive upgrade.

i honestly doubt it. the math behind it hasn't changed. and if you are winning or not is a local calculation.

if you mean "luck value" as in "blocks that made it onto the chain" - maybe. but the amount of blocks "small" miners are able to mine in theory is unaffected by network issues (you can win a block without even being online - you can't mine it, sure - but the win is the "luck" and not if it got on chain)

[NSC-FIL]

A random sampling of miners with less than 200TiB (lucky value from Filflox) seems to suggest low successful block rewards:
f0492030 (44.6TiB power) = 49.20%
f022142 (195.57TiB power) = 34.37%
f08403 (48.5TiB power) = 35%
f015927 (72.95TiB power) = 0%

I hope it's just bad luck and it all evens out over time. Can't figure out why though. :-(