Ethereum and Solana: Different Choices for Mainnet Optimization

Ethereum has begun a supply-side reform. After the dream of the infinite garden was shattered, Vitalik started to reassess the relationship with L2/Rollup and more actively defend the L1 track. The "speeding up and reducing costs" plan for the Ethereum Mainnet has been put on the agenda, with the shift to Risc-V being just the first step. In the future, catching up with or even surpassing Solana in terms of efficiency will become a key task for Ethereum.

At the same time, Solana continues to expand its consumption demand scenarios. Solana's strategy is "expand or die," steadfastly pursuing a path of large-scale L1 development. In addition to the Firedancer developed by Jump Trading entering the deployment phase, the Alpenglow consensus protocol from the Anza team became the focus of attention at the recent Solana conference in New York, attracting widespread interest from attendees.

Interestingly, Ethereum's ultimate goal is to become the world computer, and Alpenglow shares the same ambition.

New Consensus Mechanisms in the Era of Large-Scale Nodes

Since the birth of Bitcoin, the number of nodes and the degree of distribution have been considered important indicators for measuring the level of decentralization in blockchain networks. To prevent centralization, the security threshold is usually set at 33%, meaning that no single entity should exceed this proportion.

Driven by capital efficiency, Bitcoin mining ultimately evolved into a pool clustering model, while Ethereum became a major stage for some large staking service providers and exchanges. However, this does not mean that these entities can control the operation of the network. Under the model of "maintaining the network for incentives/management fees", they have no malicious motives.

However, assessing the health of a network must take its scale into account. For example, in a small group of only 3 people, a 2/3 majority is needed to be considered effectively operating; merely pursuing the minimum security of 1/3 is meaningless, as the remaining two can easily collude, resulting in a very low cost of wrongdoing and a very high benefit.

In contrast, in a large-scale network with 10,000 nodes, there is no need to pursue a 2/3 majority vote. Outside of the existing incentive model, most nodes do not know each other, and the coordination cost for collusion among large staking service providers is also too high.

So, if we moderately reduce the number of nodes and the consensus ratio, can we achieve "speed up and reduce costs"?

Alpenglow is precisely based on this idea. It plans to maintain the scale of about 1500 nodes on Solana while reducing the security consensus threshold to 20%. This not only improves the node confirmation speed, earning more Mainnet incentives for nodes, but also encourages the expansion of node scale to around 10,000.

Whether this approach will produce an effect of 1+1>2 or will break through existing security mechanisms remains to be seen. However, this line of thought aligns well with Solana's style, taking a different route from Ethereum and participating in the competition of public chains.

Alpenglow: Redefining Consensus Mechanism

The theoretical basis of Alpenglow is that in the era of large-scale nodes, a high consensus ratio is not necessary. Due to the existence of the PoS mechanism, malicious actors need to mobilize huge amounts of capital to control the network. Even at a scale of 20%, based on current prices, Ethereum would require $20 billion, and Solana would need $10 billion.

Having such a large amount of funds, choosing to control the blockchain is obviously not a wise move, and it will also face backlash from the remaining 80% of the nodes. Unless it is a state-level action, such operations are almost impossible to occur.

In practice, Alpenglow roughly divides the entire process into three parts: Rotor, Votor, and Repair. To some extent, Alpenglow is a deep transformation of the existing Turbine mechanism of Solana.

Turbine is Solana's block broadcasting mechanism, responsible for disseminating block information to achieve consensus confirmation among all nodes. Unlike the Gossip protocol used in the early design of Ethereum, Turbine adopts a hierarchical propagation approach:

1. In each cycle, the nodes are divided into Leader, Relay, and ordinary nodes, and only the Leader node can send block broadcast information.
2. A small number of Relay nodes continue to broadcast the information to more ordinary nodes, forming a Turbine Tree that resembles a tree structure.

In Alpenglow, the variant of the protocol is called Rotor, which is essentially an orderly way of propagating block messages, where no Leader or Relay node is fixed.

Votor is the node confirmation mechanism. In the Alpenglow concept, if the first round of node voting reaches an 80% ratio, meeting the minimum requirement of over 20%, it can be directly and quickly passed. If the first round of voting is between 60% and 80%, a second round of voting can be initiated, and if it exceeds 60% again, it can be finally confirmed.

If the above steps do not reach a consensus, the Repair mechanism will be activated. However, this situation is similar to the challenge period of Optimistic Rollup, and if it really comes to this, the protocol is likely to face serious issues.

Unlike simply increasing hardware resources to improve bandwidth, Alpenglow's goal is to reduce the block consensus generation process. If the data block can be kept at the current size of approximately 1500 Bytes and the generation time is short enough (with the fastest in tests reaching 100ms, which is 1% of the current 10s), then the performance improvement will be very significant.

Conclusion

After MegaETH, existing L2 solutions have basically developed to their limits. With SVM L2 unable to gain support from Solana, there is an actual demand for continued scalability on the Solana Mainnet. Only by continuously increasing the Mainnet TPS can the concept of Solana as an "Ethereum killer" be thoroughly implemented.

It is worth noting that Alpenglow is not limited to Solana; theoretically, any PoS chain, including Ethereum, can adopt this mechanism. Similar to the previously introduced Optimum, existing blockchain research has already reached the technical boundaries and urgently needs more support from computer science and even sociological concepts.

Although IBM once predicted that the world would only need five mainframes in the future, if we consider the internet built on HTTP-TCP/IP as one, Bitcoin as one, and Ethereum as one, then there is indeed not much room left for Solana. However, it is precisely this competition that drives the continuous advancement of blockchain technology, providing users with a more efficient and secure blockchain experience.