Ethereum returns to the L1 narrative, Solana consensus accelerates

Ethereum begins supply-side reform.

After the Infinite Garden dream was shattered, Vitalik restricted the debt of favor of L2/Rollup and defended the L1 track with a more proactive attitude. The Ethereum mainnet "speed up and reduce fees" plan was put on the agenda. Turning to Risc-V is just the beginning. How to catch up with or even surpass Solana in efficiency will become the next highlight.

Solana continues to expand consumer demand scenarios.

Solana’s answer is Scale or Die. It is determined to expand and strengthen L1. In addition to Firedancer developed by Jump Trading entering the deployment process, at this New York Solana Conference, the Alpenglow consensus protocol of the Anza team won the first prize and attracted the attention of the whole audience.

The ultimate dream of Ethereum is the world computer. Coincidentally, so is Alpenglow.

20% security consensus in the era of large-scale nodes

Since Bitcoin, the number of nodes and dispersion have been synonymous with measuring the degree of decentralization of a blockchain network. In order to avoid centralization, the security threshold is limited to 33%, that is, no entity should exceed this proportion.

Driven by capital efficiency, Bitcoin mining eventually moved towards mining pool clusters, and Ethereum became the main stage for Lido and CEX. Of course, this does not mean that mining pools and Lido can control the operation of the network. Under the model of "maintaining the network - earning incentives/management fees", they have no motive to do evil.

However, to measure the health of the network, its scale must be examined. For example, in a small group of 3 people, 2/3 of them must agree to be considered effective. It is useless to only pursue any 1/3 The minimum security guarantee is meaningless, and the remaining two are very easy to connect in series, the cost of doing evil is extremely low, and the profit of doing evil is huge.

If it is a large-scale network of 10,000, such as the current node scale of Ethereum given by Etherscan, there is no need to pursue a 2/3 majority vote. Outside of the incentive model, most nodes do not know each other, and the coordination cost of Lido and Binance to do evil together is too high.

If we reduce the number of nodes and the consensus ratio, can we “speed up and reduce fees”?

Everyone should be able to imagine that Alpenglow also thinks so and is prepared to do so. Maintaining Solana's 1,500 node scale and reducing the security consensus to 20% can not only increase the node confirmation speed and earn more mainnet incentives for the nodes, but also encourage the expansion of the node scale, such as expanding to about 10,000.

Whether it is the effect of 1+1>2 or it will break through the existing safety mechanism, both are possible.

But I like it very much. Solana should follow the American chain, conspiracy group, and centralized route, and participate in the public chain competition as the opposite of Ethereum.

Magic Turbine, or moving towards DPoS

Alpenglow's theoretical idea is that in the era of large-scale nodes, the number of consensus does not need to be too strong, because of the PoS mechanism, the malicious actor needs to mobilize a huge amount of capital to control it. Even if the scale is 20%, based on the current price, Ethereum needs 20 billion US dollars and Solana needs 10 billion US dollars.

With 10 billion US dollars, there are so many things you can do. Only crazy people would try to control the blockchain, and they would also face counterattacks from the remaining 80% of the nodes, unless it is a state action.

In practice, Alpenglow roughly divides the entire process into three parts: Rotor, Votor and Repair (for more detailed division and workflow, refer to its paper), and to some extent, Alpenglow is a deep transformation of the Turbine mechanism, so we will briefly introduce Turbine.

Turbine is Solana's block broadcast (Block Propagation) mechanism. Simply put, Turbine is a messenger that spreads block information to achieve consensus confirmation by all nodes.

In the early design of Ethereum, block broadcasting is the Gossip protocol, and its Chinese name is like "rumors, gossip". Any node can be used as the initial messenger, and P2P communication can be carried out between any nodes, eventually achieving the effect of full network confirmation, just like "I have a message, I only tell you, don't tell anyone", and eventually the whole world will know.

Let’s skip Turbine for now and use Ripple’s DPOS mechanism for comparison. Ethereum/Bitcoin has achieved the ultimate in point-to-point communication, so what is the cost?

It is extremely slow and expensive, which is easy to understand. The aimless and random pairing between nodes will be more robust and censorship-resistant, but it will take longer. The expensiveness is a consideration of node costs, because other nodes have slow information transmission and have to maintain their own operations, so naturally they have to pass on to users in the form of Gas Fee.

Extreme decentralization will inevitably bring about the accompanying diseases of being expensive and slow.

Ripple's DPOS went to the other extreme. In the initial stage, there were only 21 super nodes in the entire network. The extreme centralization completely solved any side effects. Now the main network nodes have reached 179.

Turbine takes an intermediate state. It neither adopts Ethereum's Gossip mechanism nor takes the evil path of reducing the number of nodes. Instead, it spreads network nodes in a hierarchical manner:

1. 1. In each cycle, nodes are divided into Leader, Relay and more ordinary nodes. Only Leader nodes can send block broadcast information.

2. 2. After receiving the information, a small number of Relay nodes continue to broadcast it to more ordinary nodes. The above process is called Turbine Tree, which is very similar to a tree, with a clear distinction between the primary and the secondary.

In Alpenglow, the variant of this protocol is called Rotor, which is essentially an orderly propagation of block messages, and any Leader and Relay nodes are not fixed.

Votor is a node confirmation mechanism. For example, in the Alpenglow concept, if the first round of node voting reaches 80% and meets the minimum requirement of more than 20%, it can be passed directly and quickly. If the first round of voting is higher than 60% and lower than 80%, the second round of voting can be started, and it can be finally confirmed if it exceeds 60% again.

If it still doesn’t work, then use the Repair mechanism, but I personally feel that this is similar to the challenge period of Optimistic Rollup. If it really comes to this, the protocol will most likely fail, and the FDIC cannot prevent the bank run on Silicon Valley Bank.

Different from brute-forcing software and hardware resources to increase bandwidth, Alpenglow’s starting point is to reduce the block consensus generation process.

If the data block is as small as possible, such as limited to the current 1500 Bytes, and the generation time is short enough, for example, in the current test,the extreme case can run to 100ms, which is 1% of the current 10s. (1 s=1000 ms, but this is questionable. If it is really used on a large scale, the conditions will not be so ideal, but it is also terrifying).

Conclusion

After MegaETH, the existing L2 has basically reached its end. After SVM L2 will not be supported by Solana, the Solana mainnet has a practical need to continue to expand its capacity. Only when the mainnet TPS overwhelms all competitors can the concept of Solana Ethereum killer be implemented to the end.

Alpenglow is not only applicable to Solana. In theory, any PoS chain, including Ethereum, can use its mechanism. Similar to Optimum, which we introduced before, existing blockchain research has reached the edge of technology and urgently needs more help from computer science and even sociological concepts.

IBM once claimed that the world would only need five mainframes in the future. If the Internet composed of HTTP-TCP/IP is one, BTC is one, and Ethereum is also one, then there is not much room left for Solana.