Does Eos's millions rate of trading per second be reliable? __eos

Maxdeath block chain postdoctoral, consensus algorithm experts 202 people agree with the answer

Before the introduction of the Dpos on the wrong, modify the answer, does not affect the conclusion.

————————————————————————————

Thanks for inviting.

The first conclusion: Of course, because EOS employs two techniques that can have infinite potential (casually blown): Dpos and fragmentation.

The finer I wrote later in the column (recently busy dragging a lot of articles), here is a simple one.

First of all, there are several numbers you first feel:

1, the transaction output of Bitcoin is 7 strokes per second, theoretically the level of transaction speed that the Bitcoin POW can support is 10 strokes per second.

2,PBFT's trading output is 1000 strokes per second, but not scalable, no one has tried more than 64 nodes will be, the basic can be considered not available.

3, the improved BFT algorithm Zyzzyva output is 10k pen per second magnitude, compared with PBFT, it is more sensitive to the number of malicious nodes, malicious nodes more time delay than pbft high. The algorithm is extensible but not infinitely extensible, that is, the number of nodes increases and the output remains unchanged, but the delay explodes.

What is the relationship between those three?

Bit-Currency system Communication Complexity O (N), can support tens of thousands of nodes, but the output can not exceed 10 strokes per second level, otherwise unsafe. (but this has been compared with the reality of the Bitcoin has a considerable gap, because the real bit currency does not have so many mining nodes, rather central, so in fact, can support higher output, here slightly not to mention)

PBFT Communication Complexity O (n^2), so not extensible, which can be considered to be strictly Byzantine fault-tolerant output caps.

Zyzzyva Communication Complexity O (N) can be extended, which is essentially the ultimate speed of broadcasting any message to the entire network.

But what is the security model for the three? The security assumption of Bitcoin is that the malicious node is less than half the total network, PBFT and Zyzzyva are less than 1/3 of the malicious nodes, but in fact, in the case of increased malicious nodes, Zyzzyva delay will be very high.

So we have a dpos,dpos security model of what it is.

The model is that evil will not be chosen again as a representative. Byzantine fault tolerance. does not exist.

In this consensus model, in fact, the consensus is particularly simple-select the individual to publish the block is good, anyway, if there is a problem will not vote for him. Communication Complexity O (N). Then, because of the centrality, there is no need to consider the transmission latency problem.

——————————————————————————————————

Bitcoin's way of reducing Byzantine fault-tolerant communication complexity to O (N) is to give cheaters punishment.

Thus, from a game theory point of view, the cost and potential benefits of malicious node cheating are simplified-the only way to benefit is double payment, which costs more than 50%.

The whole public chain consensus algorithm is in fact looking for a way to ensure that the block publishers will not cheat, or, to ensure that the cost of cheating is higher than possible benefits.

Why do you say that? Because the problem of Byzantine fault tolerance can be simplified directly to the problem of broadcast blocks as long as the block publishers are found.

Non-extensible directly becomes extensible.

Then, when everyone is looking for it in a variety of ways, demonstrating that an algorithm that all possible attacks are higher than the proceeds, Dpos appears.

It tells us--just use the votes.

What if you cheat? I'd better not choose it next time.

What is the cost of cheating? not be elected later.

How many people would be right to choose? It's good to vote.

How much is the reward suitable for? It's good to vote.

If you assume that dpos everyone is voting for themselves, Dpos is actually a POS, and the POS will be subject to an interest-free attack. If everyone is voting for a person, then this is the central system.

Then Dpos said that because everyone has the votes, so all problems are solved ...

This thing sounds very Indian.

Of course, I'm not saying that dpos must be bad, I just said that Dpos did not solve the problem, it is only to change the problem to a form. Pow afraid of mine pa, Pos afraid of no interest attack, Dpos afraid of what, can be after class exercises-attack Dpos system, need what conditions.

And if the center of voting is a center of trust, then why don't you trust the government? Or, if the center of more people's interests is the center of trust, then the central enterprise chosen by the market will meet this requirement.

Why do you think the voting system in the block chain system will be better than reality? is because in the virtual world you will be more rational than the reality. Or in the virtual world is more transparent and credible voter information.

——————————————————————————————

The above questions aside, assuming that after the Dpos, the chosen node will not cheat.

Under this premise--it's like finding a couple of nodes responsible for broadcasting all the deals.

Then, since they don't cheat, why broadcast all the deals? It would be nice if everyone believed them. Who cares about other people's trading records? So many people have bitcoin, someone with a full node.

Then, since you don't have to broadcast all the deals, why do you have each node record all the transactions? One part of each node record is not good.

Doesn't sound very reasonable. Because that's what a distributed database does.

So, is the chunk chain not a distributed database?

No, at least, not the traditional distributed database, because the traditional distributed database does not consider Byzantine fault tolerance, regardless of the malicious node.

——————————————————————————————

Fragmentation is a good technique-it can be fragmented when certain conditions are met. The academic algorithm Elastico and Omniledger all used the fragment, but, they also all explicitly wrote, when the malicious node quantity is reduced to a certain degree time, can be fragmented.

Pow,pos, or dpos can be fragmented. OK, if the proportion of malicious nodes (Calculate force/Equity/vote) is very small, because fragmentation leads to lower security.

But you can't say--because I have the vote, so the chosen person will not cheat, so we can think of it as a distributed database fragmentation, do not take this to play. Edited in 2018-01-23 202 80 comments Share collection thanks for the Jasin Yip front-end development, the excellent answer to the JavaScript topic 61 people agree with the answer to the EOS claim that millions transaction speed per second is reliable.

Whether this needs to reach this level of volume to verify that there is no rigorous mathematical and logical way to prove that it can reach this number (I hope someone can hit my face, I also learn one).

To say the digression, at the current volume level, claiming that this speed is not in addition to publicity, there is no egg to use, because the number of transactions is far from the level (Alipay 2017 double 11 per second trading peak 256,000), this should not be a high-priority job, should focus on the rest of the first place. My core question is: How to achieve a full block synchronization per second millions transaction speed.

I'm not sure what you mean by "full block sync," I'm just trying to understand that all nodes are synchronized.

In fact, EOS uses the vote to elect 21 super nodes to synchronize, only need to sync 21 nodes to complete.

It is a multicenter rather than a centralized, secure. Trust.

This problem under the low quality answer accounted for the main, which reflects the current high reputation EOS, according to the pyramid distribution, too many of the venture capital, block chain principles are not known to the people know it, so much irrational and not objective answer. We suggest that we should discuss the problem rationally, not by imagination or stereotype, but by understanding principle, technology, venture capital and sociology.

Interest Related: Cardano firm believer.

https://www.zhihu.com/question/265927960