Source: Haotian, blockchain security practitioner

During the bear market, people tend to think calmly about things that may be overlooked during a bull market, such as the resurgence of @solana in the public eye. Is the technology really viable? Can one "centralized crash" destroy everything?

Recently, MakerDAO said that Solana's codebase is good, and Visa also announced that it will pilot cooperation with Solana. Solana's TVL and other data indicators have also rebounded significantly. Is Solana going to have a second spring? How should we interpret this?

During the bull market, when faced with the soaring price of $SOL, many people praised Solana's strong capital background and community ecosystem, but few could explain the technical framework of Solana clearly.

Without exaggeration, this article provides a brief analysis of the technical architecture behind Solana from a scientific perspective, as well as the reason why Solana was not killed by Ethereum.

Tips: Only from the perspective of technical popularization, we will help you to re-understand and understand Solana, without any investment reference.

POH Consensus Mechanism

First, let's talk about the POH consensus mechanism, which stands for Proof of History. It is a special innovative mechanism adopted by Solana. In the traditional sense, there is no direct relationship between block time and physical time (weak correlation). For example, the time benchmark for public chains such as Bitcoin and Ethereum is block height, and the carrier of time passing is only the continuous stacking of block height. Therefore, the specific time for upgrades like the KanKun upgrade can only be estimated based on the network's block production rate.

Solana's POH innovation lies in anchoring the passage of time on the chain to physical time, for example: the interval time for each POH block is fixed, and continuous hash operations produce a verifiable time sequence, etc.

Doing so can avoid the manipulation and attack issues caused by "subjective" time. Imagine if the block time is irregular, it is difficult for nodes to quickly reach consensus, which can easily lead to attacks such as rollback and replay. The objectivity of physical time cannot be shaken, and nodes can monitor some abnormal situations based on the current time sequence without having to backtrack all historical data.

Therefore, the innovation of POH is to strongly bind physical time in order to promote better collaboration and consensus among Solana nodes.

In my opinion, the POH mechanism has another advantage: nodes can receive and sort a large number of transactions in batches (Pipeline), and waiting for the POH timestamp to submit transactions is equivalent to batch transactions from layer2 to layer1, this mechanism brings the idea of Rollup into layer1, providing a prerequisite for high throughput and parallel processing as mentioned below.

Storage and Computing Separation Feature

Next, let's look at the "storage and computation separation" feature, which allows Solana to overcome the bottleneck limitation of data availability.

Traditional blockchain validation frameworks rely on full nodes to publish large amounts of historical data to achieve real-time calculation of state changes. This coupling of computation and storage can limit the performance of the chain to some extent. For example, when Ethereum updates its state, it needs to synchronize all chain data and then perform historical record calculations. In addition, Ethereum is a sequential block generation process and cannot be serialized, so block generation time and single block capacity are limited.

而 Solana 将状态存储和交易执行进行了分离，有个单独的存储系统来保留状态，包括账户信息、签名者历史和交易记录等状态，有新交易执行时，Solana 会在 Pipeline 节点上高速运算，最后只更新状态到存储系统中。二者分离，可以确保账本系统快速运行，避免了 DA 校验状态时间+计算等待 DA 校验结果的时间损耗。要注意的是，网络计算和存储资源需要通过质押 SOL 获取。

Solana separates state storage and transaction execution, with a separate storage system to preserve state, including account information, signer history, and transaction records. When a new transaction is executed, Solana performs high-speed computation on the Pipeline node and only updates the state to the storage system. Separating the two ensures fast operation of the ledger system and avoids the time cost of DA verification status time + waiting for DA verification results. Note that network computing and storage resources need to be obtained through staking SOL.

To put it simply, the workers of Ethereum need to go to the warehouse to get raw materials and then go to the workshop for processing. The two responsibilities require back and forth running, which results in low efficiency. On the other hand, Solana has specialized workers in both workshops. The porters only need to move the materials that are used nearby to the production workshop from time to time, greatly improving efficiency.

High-concurrency transaction processing

Take a closer look at the "high-concurrency transaction processing" feature, which makes Solana feasible for meeting the demands of the web2 market.

Although Solana claims to have 700,000 TPS, it has been laughed at due to occasional downtime. However, MakerDAO's Endgame choice and Visa's choice are to some extent a recognition of its performance limits. So, how does Solana achieve high concurrency?

In short, it is the result of the advantages of POH and separated storage calculation mentioned earlier. To help everyone understand it more deeply, I will try to compare it with Starknet's high concurrency.

Solana received 10 transactions from user Alice simultaneously. The node will sort the transactions and wait for the POH timestamp to batch them for packaging. When the next timestamp arrives, the node will call the state data of the independent storage system to check if there are any conflicts among the 10 transactions. If there are no conflicts, the 10 transactions can be packaged into one block normally. If there are conflicts, the block with conflicts will be excluded from the packaging process.

The difference is that Alice on the Starknet network, due to the use of a unified account abstraction model, cannot have conflicting states in transactions issued by one account. Approve and Transfer can be executed simultaneously because they modify different states of the contract, with Approve corresponding to Allowance and Transfer corresponding to Balance. However, if two Transfer transactions are executed simultaneously, they will need to modify the Balance state at the same time, which can easily lead to state conflicts and calculation errors. Therefore, the account abstraction feature is the foundation of Starknet's high concurrency.

Take a simple example, Solana's parallel processing is like a restaurant that arranges multiple waiters to take orders for customers at the same time. Each waiter handles a transaction thread, and the order of ordering is also coordinated by the waiter. If there are the same dishes, the kitchen can serve them in parallel. On the other hand, Starknet's parallel processing is like using a self-service ordering machine to scan and order dishes. Customers can receive orders on multiple machines at the same time, and the backend control system will coordinate the orders.

Overall, the goal of high concurrency is to efficiently and effectively utilize system resources while maintaining network orderliness.

Solana's waiter is the node operation and maintenance system that everyone complains about for its high cost, while Starknet's self-service ordering machine is the compatible account abstraction foundation with contracts.

With a deep understanding of the underlying technology of Solana, there may be different answers to some of the criticized issues.

In my opinion, the problem with Solana lies in its overly complex technical implementation logic: 1) The high operational cost of Solana's nodes limits the number of nodes and its decentralization capabilities; 2) The POH+POS consensus mechanism requires nodes to have powerful computing and bandwidth resources to handle high concurrency, and the higher the resource cost, the greater the operational cost of the nodes; 3) There are also inevitable issues such as state conflicts and network loads when processing high concurrency.

Some people say that Solana is using web2 thinking to build a web3 network. Its technological innovation is higher than existing blockchain architectures. If we consider its problems as fault tolerance on the path of innovation, perhaps there will be a different evaluation.

The cost of innovation may be just as everyone sees it, watching it rise tall and then watching it collapse. But if the foundation of innovation remains, will this collapsed building rise again?

Note: I do not hold SOL. The above analysis is only a calm reflection and observation from a technical and business perspective. There may be some inaccuracies in the process of popular interpretation. Please take it as a popular science article and not as investment advice.

Original Link

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia