Stake Pools - extra infrastructure

I listened to a youtube video that was released today or yesterday, in this video the IOHK presenter gave some more details about staking pools - he mentioned that pool operators needed to provide more than just an “always on” basic machine, he talked about routing servers and a server farm - what’s the story? I kind of though a pool would just be an instance of SL running on a machine that was always on, now it sounds like a lot more than that… what are these routing servers? and has anyone run one?

I saw this too.
Let’s ignore the technical part first and think about a future staking pool who has to fulfil expectations like

  1. available member of a honest majority in a decentralized network of pools
  2. reliable slot leader for all occurring transactions (when elected)
  3. reliable pool for all stakers (when elected)

especially point 3 brings up an interesting question: are there any compensation rights for stakers if a staking pool fails to perform his duties?

Basic machines can suffer a lot of outages (power, connectivity, damage, attacks …) and - from my understanding - a staking-pool has to watch an entire epoch, because the slot order (or when it turns to him) is multi-part encrypted for security reasons. An epoch is announced to last 20 hours (
EDIT: …or 5 days (

Now let’s look at technical ressources: Brunjes talks about “not too less and not too much” staking pools. Too much would mean a lot more of replication and coordination overhead and so reduced throughput. But a limited number of pools has to run “the world” and for this a couple of “my old notebook” devices seems not adequate to me.

For thus who have some ISP/Datacenter/IT background it’s obviously clear that running an HA online service requires some additional things, starting by reliable and redundant equipment, going over quality and availability monitoring and alerting and not ending by motivated, educated and also redundant staff who (in our case) interestingly is working with very similar incentives to keep services running on top of “five-nines” uptime.

1 Like

It sounded like “routers” are a specific thing (not like a regular router) - a part of the way Cardano resolves SL instances or something - Does anyone have info about them?

I’m not sure if they already plan to use or emulate a RINA based communication instead of standard TCP/IP routing.
Or if they would prefer to have HA/failover frontend or a geographical distributed system of proxies.

Yes, there are different nodes in the network, even Deadauls is a node called Edge.

So, I think that a Pool needs at least one Core and one priviledged Relay node.


exactly this