The Art of Ethereum Design (Part 2)

In "The Art of Ethereum Design (Part 1)", Ethereum founder Vitalik Buterin mainly analyzed the PoS model of ETH2 in detail. Then Vitalik continued to elaborate on the expansion plan of ETH2.

Why Ethereum chooses sharding

(Note from Lvdong: The concept of sharding comes from the idea of ​​partition tables in databases. Sharding is actually about putting data on different databases and hosts so that data can be processed in parallel to improve processing efficiency. In the blockchain, each shard processes transactions within its own shard, thereby achieving the effect of capacity expansion.)

As early as 2014, Vitalik and the Ethereum community began to explore sharding. Relevant researchers initially agreed that it was stupid for every transaction on the chain to be verified by every node, and Ethereum should use a more efficient way to ensure chain security.

The study and research of computer science in the past few years has brought great inspiration to Vitalik. In the process of learning, Vitalik deeply realized that when you make an efficient algorithm a little more complicated, you can achieve the best efficiency.

A good example is the sequence sorting problem: how to sort a random sequence of numbers from small to large. Ordinary people may choose to select the smallest ones from the sequence one by one and then sort them. This algorithm is usually called O(N^2) in runtime, which means that the number of steps to process the sequence is the square of the length of the sequence.

However, this sorting algorithm is not the most efficient. A smarter algorithm may be similar to merge sort, quick sort, and some other classification algorithms. The steps required by these algorithms are no longer O(N^2), but O(LogN).

For example, if you have a sequence of 100 numbers, since 100 is a three-digit number, the required processing steps are 100*3, which is 300 steps. When you understand the operation logic of these algorithms, you will feel that these algorithms are very simple and clear from a mathematical point of view. Although these algorithms are slightly more complicated than ordinary algorithms, their efficiency is much improved.

The original blockchain was like a normal algorithm, where each node verifies each transfer. Although it is simple and easy to understand, it is very inefficient. What blockchain needs now is a slightly more complex but more efficient algorithm. Sharding may be such an algorithm. In a sharded world, validators do not need to verify everything, but only some, which is very efficient.

But how to defend against the 1% attack in this case (the attacker concentrates the computing power to attack a part of a chain and makes the attacked part unable to operate normally)? After research and exploration, researchers have now come up with several clever ways to defend against this attack.

Doubts about Sharding

There are still many people who think that sharding is not feasible. The criticisms are mainly divided into two levels:

The first is that they simply do not understand concepts such as "indirect verification" and "probabilistic verification". They believe that verification is black and white. Either you verify the transfer, or you do not verify it. There is no intermediate event.

Secondly, they have deep doubts about various assumptions in sharding, and they believe that these assumptions may make the system more vulnerable when the real system is running.

Let’s take “fraud proof” as an example. Fraud proof is essentially a large-scale “probabilistic verification”. A small group of people verify the calculation results and sign the verification proof and then stake some ETH on the proof. Anyone can verify and challenge the proof. When someone verifies the proof and finds it wrong, they can challenge it. If the challenge is successful, they can get the ETH staked on the proof as a reward.

So what is the problem with "fraud proofs"? Essentially, the problem is its "synchronous assumption". Generally speaking, when the network is operating normally, the fraud proof can be verified and synchronized to the network in a timely manner. However, when there are problems with the network and there are delays, once the "synchronous assumption" is broken, the "fraud proof" is no longer valid.

It would be best if the algorithm or mechanism can avoid these assumptions and the system can work properly without any assumptions. For sharding, researchers have been trying to reduce the number of assumptions as much as possible. For example, sharding previously relied on the "majority honesty assumption", but now even if you control 2/3 of the validators you cannot put invalid blocks on the chain. For "fraud proof", if zk Rollup is used, we can ensure the security of sharding without applying fraud proof.

Ethereum roadmap with Rollup at its core

In October this year, Vitalik published a discussion thread titled "Rollup-centric Ethereum Roadmap" in the forum. Many people wondered if Ethereum would abandon sharding and turn to the Rollup solution for expansion. Vitalik explicitly denied this statement and further explained it in detail.

(Note: Rollup is a second-layer expansion solution on Ethereum. Simply put, in order to reduce the operating burden of the first-layer main chain, Rollup will process a large number of transactions, and after the processing is completed, the results will be finally informed to the main chain.)

1. Similarities and differences between Rollup and sharding

Regarding the difference between sharding and Rollup, Vitalik said that many people now have a misunderstanding about sharding, thinking that sharding is a group of nodes, but this is not what Vitalik understands as sharding.

Vitalik believes that each shard in the shard is a logical subset of a blockchain, and then a group of nodes are dispatched to verify this logical subset, and each node can verify multiple shards.

Rollup has some of the characteristics of sharding, but not all.

One of the similarities is the splitting of on-chain computations. If you have many Rollups, different Rollups will be responsible for their own computations, which is how Rollups scale.

Another similarity is that cross-shard or cross-rollup cannot directly synchronize interaction and execution (Synchronous interaction/execution). Each shard or rollup is a separate "domain". Interactions within the domain can be synchronized, but interactions across domains cannot be synchronized. Although synchronous interaction across rollups is possible, it is very difficult.

One difference between sharding and Rollup is the security model. The security of different Rollups on a chain will be managed by the same data layer. For example, the data of Rollup on Ethereum will all be verified by ETH nodes, but sharding is not like this. So to some extent, Rollup avoids some security flaws that may occur in sharding.

Although sharding and Rollup have many similarities and differences, for users, the user experience will not be much different. Another interesting thing is that when ETH2 applies Rollup, you will find that different Rollups will use different shards. Maybe one Rollup uses 5 shards, or maybe 5 Rollups share one shard.

2. What does Rollup mean to the Ethereum community?

Vitalik first shared his views on the value difference between Bitcoin and Ethereum. Compared with the Bitcoin community, the Ethereum community is more pragmatic. For example, regarding the Ethereum Gas Limit issue, after coordination among various interest groups within the community, they will eventually weigh the best block size, rather than sticking to a unified standard, which will eventually lead to a fork.

Ethereum is now facing two very realistic problems. The surge in gas fees shows that Ethereum is in great need of expansion and hopes that expansion will be realized immediately. Excessively high gas fees have forced many applications on Ethereum to exit, especially those non-financial applications.

From a practical technical perspective, Rollup is the best existing technology option. Although state channels are available now, they are only applicable to individual applications and are too limited. Plasma is also an option, but Plasma is only applicable to the payment field and does not support general contracts. Therefore, even from a medium-term perspective, Rollup will be the best option for expansion.

Another interesting thing is that if we look at the Rollup and ETH2 roadmaps together, we can see that the PoS of Phase 0 (Phase 0), the data sharding of Phase 1, and the execution sharding of Phase 2, as long as Rollup is added to Phase 1, Ethereum will be able to achieve extremely high throughput. So as long as people are willing to continue to adhere to Rollup technology, not only can the current Ethereum achieve expansion, but in the future, the Phase 1 throughput of ETH2 will far exceed people’s imagination.

So people need to face reality, and the Ethereum community should be committed to developing Rollup. Vitalik gave several reasons:

1. Rollup can be implemented quickly and has strong scalability.

2. If the functions of the underlying public chain are simple enough, ETH2 can abandon some security assumptions such as "fraud proof", so the security will be greatly improved, which will also enable Rollup to focus more on its job.

3. As mentioned in the "Ethereum Satellite Ecosystem", if Ethereum can focus on Rollup, ETH2 will be able to cooperate with other Ethereum helpers, enablers, and even competitors in the ecosystem. Ethereum will become a platform for these projects, and these projects are like satellites surrounding Ethereum and interacting with Ethereum through bridging, ultimately creating synergy at the execution layer.

3. The impact of L2 on Ethereum’s first layer

Vitalik believes that in the future, most users will live in the second-layer network for a long time and may not interact with the first-layer network for several years. However, there may be several situations in which users still interact with the first-layer network:

1. When users can use the efficient “mass exit” mechanism to shuttle between the first and second layer networks at extremely low fees;

2. If the Layer 2 network crashes;

3. For some applications, the core of the application is more suitable to be registered on the first-layer main chain. For example, the application will issue tokens on the first-layer network, and then deposit the tokens into the second-layer network, and then people will use the tokens in the second-layer network.

Vitalik is very confident that over time, users will slowly migrate from the first layer to the second layer.

The "Ethereum roadmap with Rollup as the core" section is to be continued. In "The Art of Ethereum Design (Part 2)", Vitalik will continue to discuss the details of Rollup expansion and give an overview of EIP-1559 and Phase 1.5 of Ethereum 2.0.