How many NUCs a staker should own | ETHStaker Ethereum Validator Hardware Shopping

Transcript

So, there is something that I did want to ask. So, um how many nodes and or validators am I able to run on one of these things at once? More than you can afford. Really? Unless you are very, very rich, in which case buy a better machine. Um we've it's been tested and tested with 4,000. No problem. Um there's been a problem at 10,000. Uh the professional staking outfits stop at 1,000 for just blast radius reasons. You don't want one of those machines going down and taking more than 1,000 things down with it. So, if you have more than 32,000 ETH, buy a second one, please.

Not because you have to, but because, you know, you should. Because why not, right? Yeah, exactly. Don't forget your friend superfiz.eth. That's right. Yes, if if you want to help me reach my goal of running a thousand of those faster, mistermojo.eth, just send it all there. Yeah, yeah, yeah. All right. Well, then that means I guess for me and everybody else on the planet, for the most part, we're only going to need one, which is great because that means you just go through, you buy your one, you really don't have to worry about it anymore then.

It's great that we said that. There it's actually uh not only can one machine run a thousand validators, you should not uh buy multiple machines to run your validators. Uh so, that's Uh a lot of people love redundancy and they say, "Well, if I'm going to buy one, I I might as well buy two so that if one goes down, I can run the other one." Um in theory, there there's some some merit to that. In practice, you should never run your validator keys on two setups. It will lead to a slashing event. Um it is very serious. It is one of the most important things that we have to say to any new validator, any old validator.

Never run your keys on two machines. Under any circumstances I I I want to caveat that by like unless you know what you're doing, but if you know what you're doing, then you're not listening to me anyway. Ah, but I I I want to I want to caveat your caveat, Super. So, um early on in in mainnet, this happened. There were people who said, "Aha, but we're not just Superphiz in his, you know, basement or wherever he lives. We are smart. We are the best engineers in the industry. We can do this and we can do it safely." And it was a large staking outfit, and guess what? They got over 100 validators slashed this way.

So, that's just pure hubris. As an engineer, you know, as somebody smart, understand that you're going to screw up. So, you know, just run it in one place, and if it's down, bring it back up. That's it. there's If you want backup hardware, don't install anything on it. Leave it bare metal. Uh and that way when your system goes down, you can set it back up. Um a lot of people are, you know, they want to clone their drives and have one ready to go. Uh that's it's just dangerous. It's playing with fire. Uh so So, this is something that's near and dear to my heart because I come originally from the NAS, the TrueNAS uh community before I got into Ethereum.

And those guys built serious home servers. They cost about the same as an OC. Um they require a little bit more adult LEGO. They do not come pre-assembled. And they have a lot of this kind of redundancy built in, right? So, not the power supply isn't redundant, but how often does that fail? Um the memory is server memory. So, if it fails, it will tell you, and it will self-repair, so you can just schedule a a swap out. Um the drives are often mirrored. So, if a drive fails, just put another one in, let it re-silver, you know? Um problem solved. So, if you're really, really keen on that, I would go with just a little bit of home server redundancy in one machine rather than running two machines.

There's some merit to that. I know what I actually run servers. We we run industrial servers at our homes just because we're idiots. I mean uh We we didn't we didn't know. Not idiots. When we started this a couple of years before the beacon chain launched, we did not know what the actual hardware requirements were going to be and we didn't want to be left behind. So, Nolan and I connected because we were chatting and we're like, "Wait, we have the same server. We're both We're both fools." Um but it still it's cool. It's all right.

Uh one of the things we have run into is all right, I've run into is uh our server hardware is it's like eighth generation Intel um and it does not run all of the instruction sets required for um like go. So, uh things Some things like for example, Prism will not work on servers that are that old without a special compile. Um which is fine cuz we don't want you to run Prism right now anyway.