简单聊几个存储有关的项目：SpaceMesh、SubSpace、AO

作者： 胡飞瞳

悄悄地，加密的春天轻拂过每一个角落，区块链已在大地上扎下了深根，预示着超乎想象的繁茂。Web3的旋律愈发清晰，迈着坚定的步伐向我们走来。在这复苏的季节里，创新的种子破土而出，小草舒展着嫩绿的叶片，湖畔的花朵竞相开放，绽放着属于春天的笑容。老树在春光的滋养下，又添了一圈生长的痕迹，春笋伴随着温暖的气息，破土冲天而起。

在AI的庭院里，一切显得尤为繁荣，技术与智能交织成最美的景致。DePin草地上，生机勃勃，绿意盎然，展示着传统网络如何与 Token 共舞。这是一个充满希望与活力的春天，每一处都在讲述着创新与成长的故事，引领我们步入一个更加丰富多彩的数字新纪元。

这一轮春天的故事，是区块链向Web3转变的过程，以 Token 的交易和衍生品为基础的 DeFi 不再是唯一的主角。Web3作为新一代的互联网必然引领新一代基础设施的建设，无论是 AI 还是 DePin，讲的是新型的计算网络和存储网络，所有的 Rollup，新的共识创新，新的计算范式讲述的是新的更高性能的计算架构，具有这些基础，才有从 Web2 过渡到 Web3 的可行性。

（扯远了，收回来?）互联网的基础设施不外乎计算、网络、存储以及为其所构建的协议和应用。今天一边喝茶，一边聊聊几个与存储有关的新项目。

其实 PoC 项目并不简单，一个广义的 PoC 定义应该包括所有通过容量证明（Proof of Capacity）来达成共识的项目。正如我在[区块链随想]挖矿的三种形态：掘金(Mining)/开荒(Farming)/质押(Staking)一文中所提到的，目前能够去中心化实现PoC 验证的只有存储空间，所以我们目前看到的所有 PoC 项目都是存储相关的项目。也可以说基本上所有存储相关的项目都有 PoC 的特性。

SpaceMesh 采用 PoSt 共识机制，它能够实现存储空间承诺的验证，做到相对的公平和安全。但是，SpaceMesh 的存储空间验证不是那么严格，简单说来，它只需要两周验证一次，当然，由于 P 盘技术和速度问题，当前看起来，保存存储容量比重新P盘成本低很多，这样矿工（Farmer）更愿意一直为网络保留磁盘空间容量。

SpaceMesh 的另一个很重要的特点在于：其Plot的数据是网络协议指定的，因此不是用户数据，也可以说不是（除了共识外的）有用数据。这在 SpaceMesh 的叙事上是一个硬伤。

大致说来，SpaceMesh 的特点和发展空间可做如下考虑：

1. 无用 PoC 网络，因此，不能算是一个 Web3 存储网络。类似的项目很多，比如 Burst，CHIA等，在目前的叙事环境下不占优势，在比特币强势的情况下，没有其他叙事很难成为主流；

2. 共识简单，参与容易，比较容易形成矿池，因此比较容易有群众基础，作为一个挖矿项目而言，有其自身优势，在市场行情看涨的时候，共识容易形成并得到成长；

3. 因为其存储2周验证一次，因此SpaceMesh 对计算的需要总体比较小，也比较集中，对于计算资源比较灵活的个人和企业而言，参与成本可以比较低；但对于一般用户而言，可能计算资源会造成很大的浪费；

4. SpaceMesh 的想象空间必然要依靠 SVM（SpaceMesh Virtual Machine）支持智能合约来打开。当前 SVM 处于开发期，号称支持 WSAM。在Web3 竞争激烈的虚拟机市场，即使 SVM 研发成功，要取得市场关注将仍然充满挑战。

建议：持续关注发展，根据市场情况谨慎参与。

SubSpace 当然也是一个 PoC 项目，因为它也是依靠存储的容量证明来形成共识，但与 SpaceMesh 不同的是，SubSpace 存储的是有用数据。这个不同在叙事上带来的影响非同小可。

同时，SubSpace 的共识类似与 Nakamoto 共识，与去中心化存储龙头 Filecoin 相似，严格按照存储容量的比例，Farmer 可以自己通过存储证明来获取出块权。Nakamoto 类共识与 BFT 类共识的最大区别在于，Nakamoto 共识具有更强大的去中心化能力，基本上是可以横向扩展的，而 BFT 类共识其验证节点往往不多，达到3位数就已经非常好了。因此，Filecoin 和 SubSpace 的共识都是有用共识（采用有用存储来形成共识），同时他们也比目前大多数 PBFT 类网络更去中心化。

与 Filecoin 不同的是，SubSpace 并不是为了提供一个存储市场开放给用户使用，而是存储链历史数据，通过把链历史数据存放在多个节点来保障安全，这与比特币/以太坊等不同，比特币的全节点需要存储所有的历史数据，而不是部分。这个要求有点高，这也是比特币 SegWit 升级的原因之一，那么目前的实际情况是，大多数比特币节点不用存储所有数据，比如 SebWit 见证数据就之后少数节点存放， 以太坊由于采用的是账户状态模型，更加不用存储大量的历史数据，存储状态和近期的数据即可，因此历史数据的保存和验证是一个问题，这也是为什么 DA （Data Availability）最近大热的原因之一。SubSpace 基本上不存在 DA 问题，因为它要求整个网络存储所有历史数据，存储很多备份来保证安全。也就是说，SubSpace通过其共识机制自然地解决了 DA 问题。 

另外，SubSpace 采用共识和交易执行分离的模型，因此它是一个分层架构，在最初的设计中就做了这么一级抽象，通过 Domain 来执行交易和实现子网，实现并行计算。这一点类似 ICP，或者 Filecoin 的 IPC。
目前，SubSpace 的共识层相对完善，其合约执行，Domain交互等还有待完善。当前SubSpace 处于激励测试网的后期，有望在今年中主网上线。
建议：高度关注，有条件积极参与。

Arweave 团队近期宣布 AO 测试网上线。AO是 Arweave 之上的计算层。采用 Actor Oriented 架构。

Actor模型或Actor Oriented架构是一种并行计算的数学模型，最早由Carl Hewitt在1973年提出。这种模型旨在解决并发计算的复杂性问题，通过引入一种叫做"actors"的基本计算单位来实现。每个actor都是一个独立的实体，它们可以进行以下操作：

• 创建更多的actors：一个actor可以在运行时创建更多的actors，从而动态地扩展计算资源和处理能力。

• 发送消息：actors之间通过发送和接收消息来交互，每个消息都是不可变的，这有助于避免并发环境中的许多常见问题，如竞态条件和数据冲突。

• 处理接收到的消息：每个actor可以决定如何响应接收到的每条消息，包括改变自己的内部状态、发送更多消息给其他actors或创建新的actors。

  
  

  The author Hu Feitong introduced the blockchain to the quietly encrypted spring, which brushed every corner. The blockchain has taken deep root on the earth, which indicates that the lush melody beyond imagination is becoming clearer and clearer, and it is coming to us with firm steps. In this recovery season, the seeds of innovation have broken ground, the grass is spreading with green leaves, the flowers on the lakeside are blooming, and the smiles that belong to spring are blooming. The old trees have added a circle of growth traces under the nourishment of spring, and the bamboo shoots have broken ground with warm breath. In the courtyard, everything is particularly prosperous. Technology and intelligence interweave into the most beautiful scenery. The grass is full of vitality. full of green shows how the traditional network dances with it. This is a spring full of hope and vitality. Every place is telling stories of innovation and growth, leading us into a more colorful digital new era. The story of this round of spring is the process of blockchain transformation, and it is no longer the only protagonist based on new transactions and derivatives. As a new generation of Internet, it will inevitably lead new ones. The construction of a generation of infrastructure, whether it is about new computing networks and storage networks, all new consensus and innovations, and new computing paradigms are about new and higher-performance computing architectures. Only with these foundations can the transition from feasibility be achieved, which is far from the point of recovering the infrastructure of the Internet, which is nothing more than computing network storage and the protocols and applications built for it. The definition of "storage" should include all projects that have reached a consensus through capacity proof. As I mentioned in the article "Three Forms of Mining at Will in Blockchain", the only thing that can be decentralized and verified at present is storage space. Therefore, all the projects we have seen so far are storage-related projects, and it can be said that basically all storage-related projects have their own characteristics. By adopting the consensus mechanism, it can realize the verification of storage space commitment, which is relatively fair and safe, but the storage space is relatively safe. The verification is not so strict. Simply put, it only needs to be verified once every two weeks. Of course, due to the problems of disk technology and speed, it seems that the storage capacity saved is much lower than that of the new disk. In this way, miners are more willing to keep the disk space capacity for the network all the time. Another very important feature is that its data is designated by the network protocol, so it is not user data, so it can be said that it is not useful data except consensus. This is a difficult narrative. Generally speaking, the characteristics and development space can be considered as follows. Useless network can't be regarded as a storage network, so many similar projects, for example, are not dominant in the current narrative environment. In the case of strong bitcoin, it is difficult to become a mainstream consensus without other narratives. It is easier to form a mine pool with simple participation, so it is easier to have a mass base. As a mining project, it has its own advantages. When the market is bullish, the consensus is easy to form and grow, because its storage week is verified once, so the overall need for calculation is relatively small and relatively small. For individuals and enterprises with flexible computing resources, the cost of participation can be relatively low, but for ordinary users, the imagination space that may cause great waste of computing resources must be opened by supporting smart contracts. It is said that support is currently in the development stage. In the highly competitive virtual machine market, even if it is successful in research and development, it will still be full of challenges. It is recommended to continue to pay attention to development, participate cautiously in the architecture according to market conditions, and open the imagination space is of course a project because. It also relies on the proof of storage capacity to form a consensus, but the difference is that it stores useful data, which has a great impact on the narrative. At the same time, the consensus is similar to the consensus and the decentralized storage faucet. In strict accordance with the proportion of storage capacity, the block weight consensus can be obtained by itself through the storage proof. The biggest difference between the consensus and the class consensus is that the consensus has a stronger decentralized ability and can basically be expanded horizontally, while the verification nodes of the class consensus are often not in place. The number is very good, so the consensus of sum is a useful consensus. At the same time, they are more decentralized than most kinds of networks at present. The difference is that it is not to provide a storage market for users to use, but to store the chain historical data to ensure security by storing the chain historical data in multiple nodes. This is different from Bitcoin Ethereum, where all nodes need to store all historical data instead of some. This is also a bit demanding. One of the reasons for the upgrade is that most bitcoin nodes don't need to store all the data, such as witness data, and then a few nodes store it in Ethereum. Because the account state model is adopted, it is not necessary to store a large number of historical data, storage state and recent data, so the preservation and verification of historical data is a problem, which is one of the reasons for the recent popularity. There is basically no problem because it requires the whole network to store all historical data and store many backups. To ensure security, that is to say, it naturally solves the problem through its consensus mechanism. In addition, it adopts the model of separation of consensus and transaction execution, so it is a hierarchical architecture. In the initial design, it has made such a level of abstraction. It is similar to the current consensus layer, which is relatively perfect, and its contract execution interaction needs to be improved. At present, it is expected that in the later stage of the incentive test network, the online proposal of the main network will pay close attention to it and actively participate in it conditionally. The computing layer team on the Internet recently announced that the testing online line is the computing layer on the Internet. Architecture model or architecture is a mathematical model of parallel computing. This model was first proposed in 2000 to solve the complexity of concurrent computing. By introducing a basic computing unit, each of them can be an independent entity. They can do the following operations to create more and more at runtime, thus dynamically expanding computing resources and processing capacity. Sending and receiving messages to interact with each other is immutable, which helps to avoid many common problems in the concurrent environment, such as race conditions and data conflicts. Dealing with received messages can each decide how to respond to each received message, including changing its internal state, sending more messages to others or creating new ones. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

Actor模型具有如下的关键特点

• 并发性：Actor模型天生支持并发执行，因为每个actor都可以独立地处理消息。这使得actor模型非常适合构建分布式系统和并行应用。

• 去中心化：没有中央控制点管理actors之间的交互，这使得系统可以很容易地扩展和管理。

• 容错性：Actor模型通过隔离和消息传递机制增强了容错能力，每个actor的失败不会直接影响到其他actors，可以设计机制来监控和恢复失败的actors。

• 异步通信：actors之间的消息传递是异步的，这意味着actors不需要在发送消息后阻塞等待响应，它们可以继续处理其他消息，从而提高了效率。

看起来很美好。但是，为啥一个1973年就提出的模型，在区块链世界中现在才被 Arweave 实现出来使用呢？为啥这么多年在提升 TPS 和并发性上的努力没有考虑用 AO 架构呢？这里面的原因是多方面的，说起来，在AO 与 区块链的结合上有巨大的挑战：

1. 共识机制的限制：区块链技术的核心在于其共识机制，这需要网络中的所有参与者就某一版本的事务历史达成一致。这种全局一致性的需求与Actor模型的局部性原则相冲突，后者倾向于在分散的actors之间进行消息传递和处理。

2. 交易顺序和最终性：交易的顺序和最终性是至关重要的，因为它们直接影响到智能合约的状态和执行。AO的异步和独立特性虽然有助于提升并发处理能力，但在保证交易顺序和最终性方面面临挑战。

3. 资源和网络限制：尽管AO可以提高并发处理能力，但在分布式账本的情况下，数据同步和一致性维护的开销可能会抵消并发带来的优势。

当前我们没有看到，在采用 AO 的架构下， 如何实现整个网络的强一致性，也就是说是一条区块链；也没有看到 Actor 的验证方式，如果实现一个 Actor 执行，而在整个网络中确保其执行是正确的，不正确如何回滚等等。一个可能的实现是，基于 Arweave 的 AO 架构并不需要在 AO 计算层达成共识，他们通过 Arweave 作为 DA 来记录历史和消息，AO 之间并不需要达成一致的信任关系。因此它可能是一个弱信任网络，用于一些无需强信任的应用，比如 IM，而不是货币储值和交易等等。

建议：持续关注，技术挑战。