Ethereum's Proof of Stake consensus explained
ฝัง
- เผยแพร่เมื่อ 31 ส.ค. 2022
- An explanation of how Ethereum's new consensus method will work after the merge where it switches from proof of work to proof of stake
/ altexplainer
Donation address:
ETH/evm: 0x68655966500833332294b1984170f819Bf8b8838
Correction: 05:10, 64 subnets not 128
Correction: 05:42, double the signatures per subnet - ภาพยนตร์และแอนิเมชัน
![Proof of Work vs Proof Stake - What's The Difference? [ PoW and PoS Explained With Animations ]](http://i.ytimg.com/vi/kVuOI1IefpA/mqdefault.jpg)
![Proof of Work vs Proof Stake - What's The Difference? [ PoW and PoS Explained With Animations ]](/img/tr.png)
I made a mistake with the number of subnets per committee. There are only 64 subnets, not 128 per committee.
Also for validators who are offline, I said in the video they get slashed but it's actually an inactivity penalty which is much lower than slashing
It's so well detailed and graphically represented that I think this is the only resource that really made me understand the entire Ethereum consensus in such a short time!
Truly a gem of documentation and visualization, thank you so much for all the work that went into this video!
just to know what is your work? because all this documentations looks very professional!
Thank you, I'm glad it was helpful. An old job of mine was being the link between the dev team and clients so I used to have to try and translate a lot of technical explanations into something more understandable and would sometimes have to create documentation for it. Maybe my experience from that job is helping with the video
Man consensus mechanisms are so complex. You explain them very well here
SO MUCH DETAILED. Its one of the rare videos that had to pause multiple times and slow down to follow it with, really really appreciate it brother.
Hats off to you for the explanation and presentation. It really helped me. Thank you so much.
this video is gold !
Great job gathering so many moving pieces of the PoS mechanics in such a packed video
This is brilliant! Thank you for making such an informative, helpful video 🙌. Agreed you need many more followers.
Well made. Loved it
Thanks for taking the time and making the video!
Ethereum is brilliant in its complexity. Awesome video.
Thank you, this is the best explain of PoS I've ever seen
Amazing video, best video on PoS I found!
You really deserve more views. Fantastic work.
A question I have is how is the resulting BLS signature verified by the rest of the network? or is it only limited to each committee and the network just accepts the value as is?
Thanks for breaking it up made it easy to understand
This is simply perfect explanation. Thanks 🙏. Please do an Everscale video. I believe the top 3 technically complex L1s are Ethereum, Everscale and Cardano
I didn't know ethpos is this incredibly complex 🤯🤯
THIS WAS BRILLIANT. THANK YOU.
Holy shit bro this i fantastic. I have been reading over the paper and things got a little confusing. Really appreciate this video.
Keep making videos dude 👍
Idk if my brain expanded watching this or disintegrated.
Very good explenation!!
Wonderful explanation. I was looking for such explanations, but rarely find any article on this.
But still have some questions.
1. Why epoch is made up of 32 slots?
2. Does Stacking Smart contract holds list of all validators or just few?
3. Can the committee owner be a part of committee? or its part of another committee?
4. How many validators are there in a committee? And why only specific number?
5. What is subnets? And why 64? And how come 100 signature per subnet?
6. What do you mean by best signature? Are there any corrupts signatures? How do you define best?
Do you have any reference from where you had this information? Please share for any reference material
1. I'm not sure myself but you only need 5 digits to represent 32 in binary so it is an easy number for computers. You acheive finality every 2 epochs so they want an epoch to be completed in a reasonable amount of time. With 12 second block times, 32 slots per epoch mean you can get finality under 7 minutes which is a reasonable time.
2. The staking contract has all the validators
3. There is no committee owner. One person in a committee is chosen to create a block but they aren't the owner.
4. The number of validators in a committee is the total number of validators/32. Currently around ~14,000 per committee.
5. Subnets are just a committee being split up into smaller groups to make it more manageable. In the video I made a mistake thinking there were 128 subnets per committee when actually there are 64. The number of signatures is just the number of validators in the committee (~14,000) divided by 64. So currently there will be ~220 signatures per subnet.
I think to get the number, they looked at the worst case scenario where 100% of ETH tokens are being staked and looked at how easy or hard it would be to get all the signatures aggregated within the 12 second block time and based the number of subnets on that.
6. The best BLS signature is the one that has the most signatures aggregated within it. Some may be missing signatures.
It's been a while so I don't remember exactly where I got all the information, I watched a bunch of talks from Vitalik and other Ethereum developers and read a bunch of articles. Here are some of the sources I used:
arxiv.org/abs/2003.03052
eth2book.info/altair/part2/incentives/rewards
kb.beaconcha.in/glossary
github.com/ethereum/consensus-specs/blob/dev/specs/phase0/validator.md
lmao the best PoS explanation video on youtube has less than 1k views... I bet this will change in the next few days after the merge
This channel is a gem. Would you be interested in delving into ethereum more and maybe explain newer stuff like the sync protocol? I also noticed your channel has a very low sub number, which is a shame. Do you accept donation through patreon or on-chain? Though I’m not wealthy by any means, I really want to support!
Thank you, I am planning to cover more Ethereum. By sync protocol do you mean the sync committee that is selected to update light clients?
I just make these videos as a hobby so no need to donate. If you want to support the channel subscribing and liking the videos is plenty
@@altexplainer yes the sync committee. you showed briefly some percentage of a staker’s income is from participating in the sync committee. I’m very interested to know about how!
Brilliant . Do you have an eth address for donations to say thanks ?
great video!
beautiful
very good vedio! viewers can get a broader picture of the consensus protocal then diving into the tech materials gets much easier.Thank you. But I find it's so hard to understand your english. My first langage is not english I have to pause many many times to watch the subtitles.
thanks for the detailed explanation ! here are a few questions:
1)is "attestation" just an equivalent term to "vote", it's just the name is different, am i correct ?
2)in the video , u mentioned that in each subnet of a committee, 16 validators are chosen to produce a BLS signature. Do you mean ONLY 16 validators in each committee are chosen? or u mean each committee will be divided into many groups of 16, and then one of these groups of 16 with the best signature will be chosen?
3)for the capser part, the video mentioned that only after 2 rounds of voting might the chain of blocks be finalized. So is it just a matter of time ? coz epoch 1,2 and 3 and so on will eventually get finalized if more and more bad validators or dormant validators are eliminated, at some point of time, there will be less and less validators remaining and the 67% will be reached anyway in the end. Is this correct ?
thanks for the efforts u have put in this video !
1. Yes attestation is basically the same as a vote.
2. Only 16 random validators are chosen to produce a BLS signature. You only need 1 person to do it per committee but 16 are chosen for redundancy incase the 1 person chosen is offline or doesn't receive the signatures from everyone.
3. Yes if many validators aren't voting they will get punished by the inactivity leak so will eventually go below 16.75 ETH which is when they will be ejected from the active validator set. The validators that were voting will then make up a higher percentage of the active validator set so should be able to hit 67% again and reach finality.
Thanks, glad it's been useful
4:22 great video, just one thing i didn’t get, aren’t there 64 committees per slot, 15,000 or more validators per committee is not realistic i think
I made a mistake with the number of subnets per committee. There are 32 committees/slots per epoch and in each committee there are 64 subnets.
15,000+ validators per committee is correct though. If you look at a block explorer for the beacon chain and find a random block, it will have 15,000+ validators. E.g. I just clicked this random block 17030673 and it has 17,668 votes
beaconcha.in/block/17030673
in your cardano video you explain cardano uses a centralize universal time to know what time it is in the epoch and will try to solve that using chronos in the future. How does Eth know reach consensus on the time during an epoch?
Pretty much all proof of stake protocols all use global time servers (which aren't centralised btw but are outside of the protocol). I believe Ethereum eventually wants to switch to Verifiable Delay Functions for its randomness which will also act as clocks
@@altexplainer Thanks.
What does it mean "best BLS signature?" at 5:25.
Otherwise an amazing explanation, thank you so much.
Whichever one has the most signatures within the BLS signature. Some may be missing validators
4) in 09:52 , you mentioned that if they stopped staking , they could create a secret fork. My confusion is that why if a validator has stopped staking, they will still have the ability to add blocks to the chain to create a secret fork ?
They still have their private key that can create signatures. They can't create blocks from after they stopped staking but can create blocks from before this.
e.g. They started staking at block 100 and stopped at block 150. The chain is now on block 200. The attacker can't create a valid block after 150 but can create a valid alternate block between blocks 100 and 150. They could create this alternate block but this time in their alternate chain, not make the un-staking transaction. Then they can continue building the chain on-top of this block as in this alternate chain, they are still an active validator.
No one else who is online would recognise this alternate chain as they could see it wasn't made on time but for new nodes joining the network, they will see 2 valid chains and will need help deciding which is correct.
@@altexplainer Very well explained, very knowledgeable !!! big hugs !! But why does it have to be 51% staked for this secret fork to happen ? and how could this secret fork chain be heavier than the real chain ? And did anything like this ever happen in the past in ethereum history and how was it solved ?
This attack is theoretical for proof of stake chains. Ethereum uses weak subjectivity as mentioned in the video to prevent it from happening in the first place.
Great explanation, however one thing still eludes me: When you talk about consensus, you connect it to 32 ETH nodes. However, I do not need to stake any ETH to run my own execution/consensus client and can still participate in the consensus mechanism. 32 ETH is then required for a validator node. So is what you describe the consensus or validation process? So confusing
Yeah the naming is confusing, I agree with you there
If you are running an execution/consensus client without staking 32 ETH, you are verifying that the transactions in the blocks are correct and the signatures from the consensus method are correct but aren't actually voting in the consensus method. With this type of node you are basically keeping tabs on the block producers and stakers to make sure they aren't misbehaving.
In order to actually have voting power in consensus, you need to stake with 32 ETH
@@altexplainer So the concensus voting aka attestation is actually equal to validation? Do you know a source that would shed more light on this?
launchpad.ethereum.org/en/faq
Here's an faq on validators that might help
@@altexplainer 🙏I appreciate your help!
What do you think of the ofac compliant blocks on eth? isnt this a big problem?
It looks like the Ethereum foundation are working on a solution that should prevent any censorship so in the long term it shouldn't be an issue.
In the short term, it may be a problem if a really large percentage of validators start becoming ofac compliant but even then, even if only 1% of validators ignore ofac compliance, it just means that 'banned' transactions would take longer to be added. With 1% of validators ignoring ofac, you would need to wait on average around 20 minutes for your transaction to be included.
So I think it's still a long way away from being an issue
Why do you divide the 120,000,000 by 32ETH? Where did the numerator come from?
The total supply of ETH is 120,256,010 ETH according to coingecko
Thank you, it’s a brilliant video.
But I still have a question because I am really a beginner of eth 2.0.
I think I can not understand the process of a block from being generated to being added to the chain.
Particular, I am confused when or where the validator checks the signature of the block and also the validity of those transactions inside the block.
Below is the guess after I watched your video and read the document on the ethereum website.
After a block proposer created a block, all the validators will check the block signature and also the transactions inside the block. If they find the block is valid, they will vote the block as the latest block they think, and if not, they will vote the previous block.
Once the number of validators who think the block is valid is over 2/3 of all validators(suppose the balance of each one is the same) in this committee, the block will get the attestation and be added to the chain. if it is below 2/3, no block will be created in this slot.
If my guess is wrong, can you tell me when the signature of block and TX are validated?
You don't need 2/3 of the validators from a committee to attest to a block for it to be added to the chain. Any block created by the block proposer can be added to the chain even if it has 0 attestations.
However if it has 0 attestations, then if there is a fork and the other fork has more attestations, then using the rules from GHOST, the other fork will be the heavier chain so new blocks will build on the competing fork instead of the 0 attestation block.
Until there have been 2 epochs after the block has been added, finality is only probabilistic so there is the chance that a competing heavier fork may appear and become the heavier chain. The more attestations a block has, the less likely this is to happen though
The signature and tx in a block are validated by all nodes. If a block proposer proposes an invalid block, they can still propose the block but no honest validator will attest to it and the next honest block proposer won't add their block on top of it so it won't be a part of the heaviest chain. Even if the attacker did have 100% of the stake and got an invalid block to be a part of the heaviest chain, normal nodes that people run wouldn't accept the block as valid and would ignore the block
@@altexplainer Thank you for your explanation.
I have another question.
I think there is two round voting, and the first round we can regard it as the part of attestation from validators.
However, where do we execute the second round voting? Is the voting checkpoint(Casper FFG vote?) the second round?
Every epoch is a round of voting so the second round of voting is just waiting for another epoch
I’m confused. 120,000,000 / 32 ETH…does not equal 37.5m max nodes. Where does 120m nodes come from, anyway the math is wrong (120m/32 = 3.7m).
You are right, it looks like I accidentally used 1,200,000,000 for the calculations.
I don't think I can do anything to edit the video but I'll see if I can add a correction somewhere. Thanks for pointing it out
No problem at all just wanted to make sure I was understanding ok. Great video btw - best I've found on TH-cam. Its well paced and gives credible, and deep explanation with highly effective visuals. More please, greatly appreciated! @@altexplainer
Can you make a video explaining what is MEV? It´s only happens in ETH or other chains?
I might make a video on it in the future but it will be a while as I have other videos planned already
It's something that can happen in every chain but is more prominent with chains that have lot's of defi applications.
It's basically where someone is willing to pay block producers extra to have transactions reordered to take advantage of some sort of opportunity.
For example, maybe someone is buying a large amount of coin X which will push up the price. Maybe you also want to buy coin X but don't want to pay the new price. You can then pay the block producer a higher fee in order for them to put your transaction earlier in the block before the other person pushes up the price. The extra fee that you paid the block producer is MEV.
Whilst it is mainly seen on ETH today, if other chains start getting more opportunities for people to profit from re-ordering a block, they will also start seeing MEV. Even if the blockchain doesn't support a way to pay the block producer more on chain, people can find off-chain methods of paying block producers so it will happen on any chain where someone is able to order transactions
@@altexplainer Thanks
It seems to me that the checkpoint votes are actually the same votes as the attestations. Either I’m wrong or I can’t see this as explicitly conveyed. I can see it in a footnote in Gasper but not in Casper
Yes that is correct, the attestations from all the gasper blocks are the votes used for Casper
@@altexplainer I'm reading the Gasper protocol and still don't understand how they get to that a block has 2/3rd attestations. I don't see how it is deterministic.
Well they know how many validators there are in total for every epoch as they need to be in the staking smart contract. Right now there are exactly 866467 validators for example.
These validators are assigned to only give 1 attestation each per epoch (each validator is only in 1 committee per epoch and each committee is only allowed to vote for 1 block per epoch). If they are found to vote multiple times in an epoch, they will be slashed.
So at the end of an epoch you should be able to count all the attestations that happened and that should be the number of validators that voted for a certain epoch. If there were 800000 attestations that would be 92% (800000/866467)
Casper's rule is that if there are 2 epochs in a row where more than 67% of validators provided an attestation and it's still above 67% even if you remove any validators that got slashed for double voting then you can consider the block deterministically final.
@@altexplainer But which block would be finalized? It seems to only select one block to act as the finalized block for some epoch.
When you finalise a block in a blockchain you are also indirectly finalising all the previous blocks in the chain as that block wouldn't exist without them. So you only need to finalise the last block in an epoch to also indirectly finalise all the other blocks before it
11:26 why did the block with only 10000 attestations become the most attested one?
Because it was competing with the block below it that only had 2500 attestations
@@altexplainer thought it was competing with others horizontally
I don't understand why one epoch is not enough for finalization. If one third (malicious) vote for 2 different epoch, then only one of the epoch will have more than 2/3 of the votes. In your example, 4 out of 10 nodes are malicious. This is more than one third and and incorrect
It depends on the level of finality you want
After 1 block, there's already pretty much a 99.99999% finality if you assume an attacker has less than 20% of the stake
After 1 epoch, there's finality if you assume an attacker has less than 33.333% of the stake
For finality even if an attacker has more than 34% stake, you need to wait for 2 epochs in a row with more than 67% stake and less than 33% of the stake slashed before you can be certain that the block is final
Why would a person with 51% stake do LONG RANGE attack instead of the 51% attack?
With the long range attack, you can buy 51% of the stake, stake it, stop staking it, sell it, then start the attack. This means when you are actually doing the attack, you will have no exposure to Ethereum so there's no risk and in theory could do the attack for no cost if you were able to buy and sell at the same price.
Also in another way to do it, you don't even need to ever actually buy eth as you could try buying the private keys of validators who have already stopped staking. You need access to the private keys from 51% from any point in time. You don't actually need the 51% right now like the 51% attack
Long range attack they are not risking any capital and with a 51% attack they are risking it all. There is social slashing and defenses on chain against 51% attacks. They would lose there entire stake and gain almost nothing. If i can find the article Vitalik wrote about it ill post it.
Hi, there is a interesting protocol called Prism that claims to be able to scale Bitcoin by 10000X. Can you make a video on it?
Currently you can run a bitcoin node on really lightweight hardware like a raspberry pi and this makes running a bitcoin node very accessible which makes it very decentralised.
The problem with Prism is that it requires nodes to be a lot more powerful in order to achieve that throughput.
If you increase the hardware requirements for nodes you are sacrificing decentralisation. Bitcoiners don't want to sacrifice decentralisation so Prism will never be implemented on Bitcoin.
I probably won't make a video on Prism as I don't think bitcoin will ever implement it but Cardano is implementing input endorsers which work similarly to Prism and I was thinking about making a video on that
@@altexplainer But hardware-wise excluded, how it really works?
@@altexplainer Also, can you share with us the secret of being able to understand so many complicated topics without being in the blockchain industry? You just read the whitepaper?
th-cam.com/video/gTJyDtuWvUQ/w-d-xo.html
This is a good video that goes through how it works. Essentially they split the block up into smaller parts and have a continuous stream of smaller blocks.
Yeah reading whitepapers and getting into arguments on discord or twitter
Who's the ghost chain now?
Is etc better than eth?
ETC uses proof of work and isn't used anywhere near as much as ETH. One of the main things that add value for cryptocurrencies is network effects. The more people who use a crypto, the more valuable it becomes. ETC isn't really being used so isn't better than ETH.
those the reward for validators and proposers come from the gas fees? or is new Ether being minted? and if so, whats the rate of minting, and whats the rate of burning? is Eth inflationary or deflationary? I cant find any information about what happens with the gas fees other than whats said in EIP-1559. Does it still behave the same? so we are to assume the reward per block is the percentage minted to the proposser, and the percentage minted to the validators, plus the priority fee all to the proposer?
eth2book.info/capella/part2/incentives/rewards/
This is a good resource for more in-depth areas of Ethereum staking.
The reward for validators comes from newly minted Ether but it's hard to predict the exact rate of minting as the more validators in the network, the more Ether gets minted.
The burn rate still works the same as EIP-1559 and is also difficult to predict as it comes from how much users are willing to pay in transaction fees. During bull markets there will be more fees burned and during bear markets there will be less.
thanks for the detailed explanation ! here are a few questions:
1)is "attestation" just an equivalent term to "vote", it's just the name is different, am i correct ?
2)in the video , u mentioned that in each subnet of a committee, 16 validators are chosen to produce a BLS signature. Do you mean ONLY 16 validators in each committee are chosen? or u mean each committee will be divided into many groups of 16, and then one of these groups of 16 with the best signature will be chosen?
3)for the capser part, the video mentioned that only after 2 rounds of voting might the chain of blocks be finalized. So is it just a matter of time ? coz epoch 1,2 and 3 and so on will eventually get finalized if more and more bad validators or dormant validators are eliminated, at some point of time, there will be less and less validators remaining and the 67% will be reached anyway in the end. Is this correct ?
thanks for the efforts u have put in this video !
Yes this is correct. All consensus methods are basically ways to vote