Content
@
0 reply
0 recast
0 reaction
Lefteris Karapetsas
@lefteris.eth
😬 If client diversity is so important for ethereum can we get more accurate monitoring tools that are not based on self-reporting and provide confidence on the displayed stats?
7 replies
11 recasts
94 reactions
Izzy 🎩
@izzyp
There's a bunch of different efforts in this direction, but ultimately it's actually a very hard (technical) problem. There's no way to get confidence on the data unless you have control over operator systems (and no one wants that) or you're running everything in a TEE (won't happen from technical+speed perspective).
2 replies
0 recast
5 reactions
Lefteris Karapetsas
@lefteris.eth
I know it's a hard technical problem but it needs to be addressed. We can't fly blind as a community. Also the systems monitoring diversity should be air-tight. Not like this one here where it does not even add properly up to 100%. Gives zero confidence on correctness.
1 reply
0 recast
4 reactions
Izzy 🎩
@izzyp
There's a lot of nuance here that you're sidestepping. Data doesn't end up to 100% for a variety of logical reasons, here's one example: a validator client can be connected to multiple EL nodes at the same time (via various manners). What's the correct way to report this?
1 reply
0 recast
0 reaction
Lefteris Karapetsas
@lefteris.eth
just report the primary not secondary EL. The point of these statistics are to guide people to decide which EL/CL to run. Which means you only care about the primary EL, as in the armageddon case this is what would affect how the network behaves
1 reply
0 recast
0 reaction
Izzy 🎩
@izzyp
It's not always that clear. For example: (this applies to maybe like only 10% of all operators, but they run a large % of validators) with Vouch you can bypass "primary" EL for specific duties (e.g. attestations), or you have multiple ELs and just pick the one that responds the fastest, or dual EL consensus (nimbus).
1 reply
0 recast
2 reactions
