It’s currently a problem that we do not have more relays, especially from organizations in differing geographies. But as it stands relays are not cheap and there is no incentive to run one unless you’re also a builder.
- Crowdfund and create grants for new relays? Could also work with all current relays to open source knowledge on operation and infrastructure. It would seem in the best interest of validator pools to fund such grants.
- Normalize relays taking profit? This would most likely require unilateral agreement on fees to charge as the most stable equilibrium is 0% right now.
- Use relay enforced builder staking to generate yield.
- Other ideas?
Any resource about the cost of running a relay?
Service-wise, for a redundant and scalable setup, you’ll probably want these:
- Load balancer + Firewall
- 2x proposer API (4 CPU, 1GB RAM)
- 2x builder API (2-4 CPU, 1GB RAM)
- 2x data API (1 CPU, 1GB RAM)
- 2x website (1 CPU, 2GB RAM)
- 1x housekeeper (2 CPU, 1GB RAM)
- Redis (1GB)
- Postgres DB (100GB+)
- A bunch of beacon-nodes (3 for redundancy?)
- Block validation EL nodes
So based on that set up its around $20k USD per year, which wouldn’t include networking costs or a staging environment. On top of that operators would most likely expect some type of compensation so 2-3x’ing the cost to $40-60k doesn’t seem outlandish.
I’m not too up to date on DAO treasury grant sizes, but this certainly seems feasible!
Additional ‘fully loaded’ costs to consider:
- Ongoing maintenance – non-trivial in an emerging/rapidly evolving category
- Driving (and maintaining) validator adoption – i.e. the ‘build it and they will come’ fallacy
Would add to that the cost of 24/7 devops coverage in case there are any errors
Fwiw, I’ve updated the RAM requirements in the post above, which significantly dropped after recent updates to flashbots/mev-boost-relay.
I just noticed this part from watching video,so will relay still exist if PBS is alive,if builder and proposer can’t conduct a block bypass PBS, what’s the meaning to incentive relay?
We effectively have opt-in Proposer-builder separation today with MEV-Boost. The interactions between proposers and builders is and performed off-protocol thought and requires a trusted relay inbetween.
After in-protocol PBS has been implemented these relays will no longer be needed. This development may take several years though and until then we need a healthy diversity of relays.