Hierarchical erasure code storage method for MPT partition scene

Abstract

The invention discloses a layered erasure code storage method oriented to an MPT partition scene, which is characterized in that a method of introducing a coding and decoding module into each copy node is adopted, updated state nodes are coded and distributed through layered erasure codes, decoding is executed according to needs under a read request or state recovery scene, and the storage efficiency is improved. The method realizes state storage and reading in a decentralized environment, and comprises the steps of constructing a P2P cluster network, determining a data storage structure, organizing and dividing nodes, encoding data blocks, performing distributed storage and reading and the like. Compared with the prior art, the method has the advantages that MPT distributed storage with high fault tolerance and high efficiency is achieved, the expandability of the system is remarkably improved, the problems of high data dependence, high node repair cost, high cross-fragment recovery overhead and the like in a fragment state are solved, the method is particularly suitable for the situation that a large-scale MPT key space or a sub-tree range is divided into a plurality of independent partitions, and the method is suitable for popularization and application. And distributed storage and reading of data blocks and redundant blocks in a decentralized environment are realized.

Description

Technical Field

[0001] This invention relates to the fields of state partitioning and distributed storage technology, and in particular to a hierarchical erasure coding storage method for MPT partitioning scenarios. Background Technology

[0002] In account-based blockchain systems, system state is typically organized and stored in the form of a Merkle Trie (MPT). The MPT maps account addresses to account states, maintains key-value index relationships through a tree structure, and utilizes a hash mechanism to verify the integrity of state data. At each block height, the execution of a transaction triggers the reading and modification of several nodes in the MPT, forming a new state root hash to represent the latest state of the system. To ensure decentralization and security, existing blockchain systems generally adopt a full-replication state storage model, where each replica node maintains a complete copy of the MPT state tree. Under this model, each node can independently perform state verification and queries. However, as blockchain transaction throughput and state update frequency increase, the number of nodes in the MPT expands, causing individual nodes to bear ever-increasing storage overhead. This full-replication model results in state storage costs increasing linearly with the number of nodes, making it difficult to meet the scalability requirements of large-scale systems.

[0003] To address the aforementioned issues, distributed systems often employ data partitioning or erasure coding techniques to reduce the storage burden on individual nodes. Erasure coding, by encoding the original data (i.e., MPT nodes) into data blocks and redundant check blocks, enables the system to recover the original data even if some data is lost. Some research has attempted to introduce erasure coding into blockchain systems, combined with Byzantine fault tolerance mechanisms, to address scenarios where malicious nodes intentionally discard or tamper with data. However, most existing methods primarily optimize block data, which typically exhibits linear, append-only characteristics, making its encoding and recovery methods difficult to directly apply to MPT data with its hierarchical structure and node dependencies. Unlike block data, MPT, as a tree structure, has explicit parent-child relationships and path dependencies between its nodes. Distributed storage of MPT nodes requires not only encoding and distributing node content but also ensuring the correct recovery of the structural relationships between nodes and supporting deterministic access based on key paths. Furthermore, during state reading, the client needs to traverse from the root node to the target leaf node, a process that typically involves accessing multiple storage nodes, resulting in high communication overhead. When some replica nodes become unavailable due to faults or malicious behavior, the system typically needs to recover the missing MPT nodes through erasure coding decoding mechanisms.

[0004] However, existing erasure coding schemes mostly employ coarse-grained block encoding. Even when only a single MPT node needs to be recovered, it often requires decoding the entire block containing multiple nodes, resulting in a decoding amplification effect, increasing computational costs and access latency, and failing to meet the performance requirements of high-frequency state reads. On the other hand, in a Byzantine environment, erasure coding itself only guarantees data recoverability, not the correctness of the recovered data. Malicious replica nodes may return tampered MPT node data; without an effective verification mechanism, this will affect the correct reconstruction of the state tree. Although MPT natively relies on hash mechanisms for integrity verification, how to efficiently verify recovered data at the fine-grained node level in distributed storage and erasure coding recovery scenarios remains a pressing issue.

[0005] In summary, existing MPT-based state storage technologies suffer from high storage overhead in full-replication mode, low recovery and verification efficiency in Byzantine environments, strong data dependencies, high node repair costs, and significant cross-shard recovery overhead in sharded states. Therefore, how to achieve efficient distributed storage of MPT nodes while maintaining the verifiability and indexability of the MPT structure, reducing the storage overhead of a single node, and realizing fine-grained, low-overhead node recovery and verification in a Byzantine fault-tolerant environment has become a significant technical challenge in the current blockchain state storage field. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a hierarchical erasure coding storage method for MPT partitioning scenarios. This method employs hierarchical coding and sub-block-level recovery to achieve efficient recovery and verification of state nodes, reducing single-node storage overhead while maintaining Byzantine fault tolerance and high-performance state retrieval requirements. Specifically, MPT partitioning refers to dividing nodes or subtrees within an MPT into multiple independent node subsets according to predetermined deterministic rules. Each node subset contains one or more interconnected MPT nodes, serving as the basic unit for subsequent data block organization, erasure coding, and distributed storage. This method, through partitioned storage and hierarchical coding of MPT state data, significantly reduces the storage burden on single nodes while maintaining high-performance state retrieval, ensuring state recoverability and verifiability. This invention applies erasure coding to subsets of MPT nodes at different levels, enabling the state recovery process to proceed incrementally step by step. The lower layers use low-redundancy coding parameters to support local, lightweight state node recovery, while the higher layers use high-redundancy coding parameters to provide final data availability guarantees when lower-level coding fails to recover the state. This achieves a balance between recovery cost and fault tolerance. To further reduce computational and communication overhead during state recovery, this invention introduces a sub-block-level recovery mechanism during erasure coding. This involves further dividing the coding unit into fixed-size sub-blocks. During state reading or recovery, decoding is performed only on the sub-blocks related to the target state node, avoiding the overall decoding of data blocks containing multiple nodes. This reduces decoding amplification effects and lowers network transmission costs. In state reading scenarios, this invention maintains the original index and path access semantics of MPT, allowing clients to still perform deterministic access to the state based on key paths. When encountering missing nodes or unverifiable node data during access, the system triggers erasure coding decoding, progressively escalating the recovery level according to a hierarchical coding strategy until the required MPT node data is successfully recovered, thus ensuring the continuity and correctness of state reading in a Byzantine environment. Furthermore, to compensate for the limitation of erasure coding in guaranteeing only data recoverability but not data correctness, this invention introduces a fine-grained integrity verification mechanism. This mechanism verifies the encoded state nodes and their redundant data, enabling independent verification of the recovered MPT node data and preventing malicious replica nodes from returning forged or tampered data. Under the same fault tolerance assumptions, this invention reduces state storage overhead by up to 92.3% while still supporting high-concurrency state reading and efficient recovery, significantly improving system scalability and practical application value.

[0007] The specific technical solution to achieve the purpose of this invention is: a hierarchical erasure coding storage method for MPT partitioning scenarios, characterized by the use of hierarchical coding and sub-block level recovery methods to achieve efficient recovery and verification of state nodes. This reduces single-node storage overhead while simultaneously considering Byzantine fault tolerance and high-performance state reading requirements. The specific execution flow of this method includes the following steps: Step 1: Construct a P2P cluster network, configure node identity information, and start the cluster machines. The cluster is a collection of server nodes in the network that perform the same duties. The node identity information includes node role, node IP, node ID, and public and private keys. The node role includes consensus node and execution node. The node IP is used to represent the IP address of the server node. The node ID is used to uniquely identify the server node. The public and private keys are used for encrypting messages and authentication.

[0008] Step 2: Each node generates a defined MPT-based data storage structure through a consensus mechanism, and confirms the consistency of the MPT-based data storage structure through the consensus mechanism to ensure that each node holds the same MPT structure; after consensus is reached on the MPT data storage structure, each node enters the data submission stage.

[0009] Step 3: During the data commit process, the encoding module in each replica node is triggered to collect the MPT nodes generated during block execution and organize these nodes into multiple minimum data blocks. Each node needs to explicitly identify and collect the MPT nodes that have been updated in the current commit phase.

[0010] After completing the collection, the encoding module prunes the MPT according to preset deterministic pruning rules to divide the updated MPT nodes into multiple node subsets. The deterministic pruning rules are used to specify the pruning position and node affiliation. The deterministic pruning rules may include a method of allocating nodes according to the number of tree layers. When pruning forms a parent-child boundary, the parent node located at the pruning boundary records the index information of its corresponding child node, so that even after a node is divided into different node subsets, the deterministic location and access of the child node can still be completed based on the index information. After completing the node subset partitioning, the encoding module performs data block processing on the tree node data within each node subset. Specifically, it concatenates multiple tree node data belonging to the same node subset, including the data of each tree node itself and the index information of its possible child nodes, in a predetermined deterministic order to form at least one data block. The index information is used not only to record the structural relationship between parent and child nodes but also to record the storage location information of the corresponding tree node data in the data block. The storage location information includes at least the data block sequence number to which the tree node data belongs, as well as the offset position and data length of the tree node data in the corresponding data block, so that the corresponding tree node data can be deterministically located and extracted from the target data block based on the index information during subsequent reading, verification, or recovery processes.

[0011] Step 4: After generating the data blocks, the encoding module encodes them using a multi-level hierarchical erasure coding method. The lower-level encoding supports lightweight incremental recovery, while the upper-level encoding ensures data availability in the event of Byzantine faults. Specifically, multi-level hierarchical erasure coding enables different levels of redundancy at different levels, allowing the same data block to have varying degrees of redundancy protection at different levels. At lower levels, the encoding module uses the data block or its sub-blocks as the encoding object, generating redundant check data using a relatively low-redundancy erasure coding method. This supports local, lightweight recovery in scenarios with limited data loss or corruption. Through the lower-level encoding, data recovery can be completed without triggering a high-overhead full recovery process when only some replica nodes fail or data is lost. At higher levels, when the lower-level erasure coding mechanism cannot complete data recovery, the encoding module progressively enables higher-redundancy, more fault-tolerant erasure coding levels to form global protection for the data block.

[0012] The upper-level coding is activated in a tiered triggering manner. That is, after a recovery attempt at a lower level fails, the redundancy strength and fault tolerance stringency of the erasure coding are increased layer by layer according to a predetermined hierarchical order until data recovery is complete. At the highest level, the erasure coding level satisfies the system's preset fault tolerance assumptions, ensuring that even in scenarios involving multiple replica node failures or malicious nodes intentionally discarding or tampering with data under Byzantine fault conditions, the data block can still be successfully recovered, thereby maintaining system data availability.

[0013] Furthermore, to further reduce the computational and communication overhead during the decoding process, the encoding module further divides each data block into multiple fixed-size sub-blocks, using these sub-blocks as fine-grained units for erasure coding and decoding. Sub-blocks from different data blocks with the same sub-block number together form a stripe. This stripe serves as the smallest atomic unit for performing erasure coding decoding. During data recovery, the system determines the stripe range overlapping with the MPT node content based on the offset position and data length of the MPT node to be recovered within its respective data block. Erasure coding decoding is then performed only on the corresponding sub-blocks within this stripe range, thus avoiding the reconstruction of the entire data block. This enables on-demand recovery based on sub-blocks, significantly reducing bandwidth consumption and computational overhead during data recovery.

[0014] To ensure the verifiability of the data blocks and their corresponding redundant check blocks in distributed and Byzantine environments, this invention further introduces a Vector Commitment (VC) mechanism to uniformly commit to the data blocks and redundant check blocks. Specifically, the encoding module assigns a unique data block sequence number to each data block and, based on the hierarchical information of the data block and its corresponding data block sequence number, assigns corresponding sequence numbers to the redundant check blocks generated by erasure coding at different levels. Subsequently, the data blocks and redundant check blocks are organized as vector elements according to the order determined by the sequence numbers, and the vector elements are committed to through vector commitment. To ensure the verifiability of the recovery results based on sub-blocks, the encoding module introduces a Merkle tree-based integrity verification mechanism for the sub-blocks within each data block. Specifically, the encoding module constructs a corresponding Merkle tree by using multiple sub-blocks belonging to the same data block as leaf nodes and calculates the Merkle root digest of the data block. The Merkle root digest is used to characterize the integrity status of all sub-blocks within a data block. The generation rules for the data block sequence number and the redundancy check block sequence number, the arrangement order of the vector elements, and the generation process of the vector commitment are all deterministic, ensuring that different replica nodes can obtain consistent vector commitment results when independently executing the above process under the same input conditions. Through this method, any data block or redundancy check block can be independently verified using its corresponding commitment proof, thereby guaranteeing the integrity and correctness of the data block and redundancy check block during distributed storage and recovery.

[0015] The node subset and the redundant blocks generated by erasure coding are organized as vector elements in a predetermined deterministic order, and the vector elements are committed to to ensure the integrity and verifiability of the node subset and redundant blocks. During state reading or state recovery, based on the storage location of the target MPT node data in its data block, erasure coding decoding is performed only on the sub-blocks that overlap with the target MPT node data to recover the target MPT node data as needed.

[0016] Step 5: Distribute and store the encoded data blocks and their corresponding redundant check blocks to multiple replica nodes in a decentralized manner. This allows different replica nodes to hold a portion of the data blocks or redundant check blocks. The decentralized distribution involves each replica node negotiating the encoded data blocks and redundant check blocks using a deterministic protocol. This deterministic protocol determines the data blocks or redundant check blocks each replica node needs to store based on predetermined allocation rules, without requiring a central coordinating node. The deterministic protocol can be executed as an independent allocation protocol or based on a corresponding consensus protocol under different trust assumptions. The consensus protocol includes, but is not limited to, Byzantine fault-tolerant consensus protocols or crash fault-tolerant consensus protocols, aiming to ensure that all replica nodes in the network reach a consensus on the final allocation method of the data blocks and redundant check blocks. After executing the deterministic protocol, each replica node retains only the data blocks or redundant check blocks allocated to it and deletes data unrelated to its own node, thus enabling different replica nodes to hold a portion of the data blocks or redundant check blocks, completing the distributed storage of the data blocks and redundant check blocks. Each replica node retains data including data blocks stored by the replica node and corresponding redundancy check blocks. Each data block includes at least data from one or more MPT nodes and index information of their child nodes associated with the MPT node. Additionally, each replica node stores the commitment value and verification proof of the vector commitment corresponding to its retained data blocks or redundancy check blocks, used to verify the integrity and correctness of the data blocks or redundancy check blocks.

[0017] Compared with existing technologies, this invention achieves a balance between low storage overhead and efficient state access while ensuring the recoverability, verifiability, and Byzantine fault tolerance of state data. Furthermore, it achieves near-optimal redundancy configuration while maintaining constant storage overhead for each replica node. Experimental results show that, under the same fault tolerance assumptions, this invention can reduce state storage overhead by up to 92.3% while still supporting high-concurrency state reading and efficient recovery, significantly improving the system's scalability and practical application value. Attached Figure Description

[0018] Figure 1This is a system diagram of the architecture of the present invention; Figure 2 A flowchart illustrating the data block composition and generation process; Figure 3 A schematic diagram of layered erasure coding and verifiable storage; Figure 4 A schematic diagram of the process for selecting a data block replica. Detailed Implementation

[0019] This invention effectively reduces the storage overhead of a single replica node by using hierarchical encoding and distributed storage of state data, while ensuring the recoverability and verifiability of the state data. The method employs hierarchical erasure coding and fine-grained recovery mechanisms, prioritizing low-overhead data recovery methods in common failure scenarios and providing data availability guarantees that meet system fault tolerance assumptions in extreme failure situations. Simultaneously, by introducing a verifiability mechanism to uniformly verify stored data, the system can maintain the correctness and consistency of state data even in distributed and Byzantine environments, thereby improving the system's storage efficiency and scalability.

[0020] The present invention will be further described in detail below with reference to the specific embodiments and accompanying drawings. Except for the specific details mentioned below, the processes, conditions, and specific implementation methods of this invention are common knowledge to those skilled in the art, and the present invention does not impose any particular limitations on them. Example

[0021] See Figure 1 Without affecting the understanding of this invention, aside from processes less relevant to this invention such as client transaction submission and consensus ordering, the core processes of this system mainly include an encoding module and a decoding module deployed in each replica node. The encoding module takes the MPT state structure formed after block execution as input, is responsible for collecting updated MPT nodes and organizing them into data blocks, and encoding the data blocks using a multi-level hierarchical erasure coding method. The data blocks and redundant check blocks output by the encoding module are distributed and stored in multiple replica nodes in a decentralized and verifiable manner.

[0022] In the layered erasure coding, the lower-level coding is used to support lightweight, incremental data recovery, while the higher-level coding is used to provide final data availability assurance in the presence of Byzantine nodes. The decoding module takes the client's status query request as input and is responsible for traversing the MPT structure from the data blocks stored in the replica nodes to obtain the target status data. When the target data is unavailable or fails verification, the decoding module takes the data block and redundant check block as input, triggers the layered erasure coding recovery process, and attempts to decode level by level in hierarchical order until complete and verifiable MPT node data is output, and returns the data as the status reading result to the client.

[0023] The data block construction and generation process based on the MPT structure addresses the issue of explicit parent-child dependencies between nodes in the MPT. Each parent node must contain indexes and location information for the data blocks containing its child nodes. Therefore, nodes in the state tree cannot be serialized and assembled into data blocks in an arbitrary order. This invention organizes MPT nodes in a bottom-up manner to ensure the correctness and integrity of node relationships within the data blocks. In this invention, nodes participating in the construction of data blocks can be divided into two categories based on their positional relationship within the MPT: one category consists of nodes that do not yet contain child node reference information, and the other category consists of nodes whose child nodes have all been placed and can generate complete metadata.

[0024] See Figure 2 The process of constructing and generating data blocks based on the MPT structure is as follows: 1) The encoding module starts processing from the leaf nodes of MPT, serializes the leaf nodes and puts them into the corresponding data blocks. Since the leaf nodes do not depend on other nodes, they can be directly inserted into the data blocks and placed first.

[0025] 2) After the leaf node is placed, the encoding module generates corresponding metadata information for the leaf node and embeds the metadata information into its parent node, so that the parent node can record the index information of the data block where its child node is located.

[0026] 3) Once all child nodes of a parent node have been placed and have complete index information, the encoding module serializes the parent node and places it into a data block.

[0027] 4) The above process is executed recursively from bottom to top along the MPT structure until the root node is processed, thus completing the data block layout of the entire state tree.

[0028] See Figure 2Leaf node L1, as the terminal node of the state tree, is first serialized and placed into a data block. After L1 is placed, the system generates its corresponding metadata information and embeds this metadata information into its parent node B0. Subsequently, parent node B0 is serialized and placed into a data block. The above process continues recursively upwards until the root node of the state tree is processed, thus ensuring that when any parent node is serialized, all its child nodes have been placed and have complete reference information. After the data block structure is completed, to meet the requirements of erasure coding, data blocks belonging to the same erasure coding group are uniformly adjusted to an equal-length structure. When the actual lengths of different data blocks are inconsistent, the shorter data block is padded. At the same time, each data block is further divided into multiple sub-blocks of fixed size. The sub-block serves as the basic unit for erasure coding and decoding. Sub-blocks from different data blocks that have the same sub-block number together form a stripe. The stripe serves as the smallest decoding atomic unit, enabling the system to recover data from a single or a small number of nodes without reconstructing the entire data block, thereby reducing the computational and communication overhead during the data recovery process.

[0029] See Figure 3 As shown in the figure, the hierarchical erasure coding and verifiable storage are implemented. Each replica node collects and organizes updated MPT node-generated data blocks, and uses a multi-level hierarchical erasure coding method to generate redundant check blocks to support subsequent data recovery. For example, in a lower coding level, each coding group contains K_1=2 data blocks and generates M_1=1 redundant check blocks, enabling the system to complete recovery even if any data block is missing. When the available data in this level is insufficient to complete recovery due to replica node unavailability or malicious behavior, the system upgrades the recovery process to a higher coding level, jointly coding all data blocks to tolerate more missing data blocks and complete recovery. To ensure the verifiability of data blocks and redundant check blocks, this invention introduces vector commitment (VC) at the data block level for commitment and verification. When the target replica node R_target stores data block C_i, it generates a vector commitment φ^h during the data commit phase and calculates and saves the opening proof Π_i of the data block in the vector commitment. When a client requests or recovers data block C_i through erasure coding, the target replica node returns data block C_i and the corresponding enable proof Π_i. The client uses the enable proof to verify whether data block C_i is correctly bound to the position at index i in the vector commitment φ^h, thereby confirming the validity of the data block.

[0030] Furthermore, a Merkle tree integrity verification mechanism is introduced at the sub-block level: multiple sub-blocks belonging to the same data block are used as leaf nodes to construct a Merkle tree, and its root is used as the commitment θ of the sub-block. The target replica node provides the corresponding Merkle proof π_(i,j) along with the returned sub-block. The client uses the Merkle proof to verify whether the sub-block D_(i,j) belongs to the data block C_i. Only when both the Merkle verification of the sub-block and the vector commitment verification of the data block pass are the sub-blocks accepted as valid data.

[0031] Referring to Figure 4, the data block replica selection process based on a deterministic protocol is shown. To ensure consistency in the data block allocation and storage process, this invention introduces a deterministic protocol during the data submission phase. This enables each replica node to reach a consensus on the set of replicas participating in data block storage at the same block height, thus providing a unified basis for subsequent data block distribution, storage, and recovery processes. For example, taking the PBFT consensus protocol as an example, as shown in the figure, during the submission phase at block height h, after completing consensus, each replica node, based on the consensus result formed during the consensus process, jointly determines the set of replicas Rh used for data block storage at the current block height according to deterministic rules. In this example, after the consensus submission is completed, each replica node sends a signed confirmation message to the consensus leader node. When the leader node collects no less than a predetermined number of valid confirmation messages, it summarizes the confirmation messages and writes them into the block header of subsequent blocks. As subsequent blocks are finally confirmed, each replica node can determine the same set of replicas R_h from the consistent block header information and perform the allocation and storage operations of data blocks and redundant check blocks accordingly. Using the above methods, the system can still guarantee the global consistency of the replica selection results even in the presence of asynchronous network or Byzantine nodes.

[0032] The scope of protection of this invention is not limited to the above embodiments. Any variations and advantages that can be conceived by those skilled in the art without departing from the spirit and scope of the inventive concept are included in this invention and are protected by the appended claims.

Claims

1. A layered erasure code storage method for MPT partitioned scenarios, characterized in that, The method of introducing encoding and decoding modules in each replica node is adopted to encode and distribute the updated state nodes by hierarchical erasure code, and to perform decoding on demand in the read request or state recovery scenario, so as to realize the state storage and reading in the decentralized environment, and the method specifically comprises the following steps: Step 1: Construct a P2P cluster network, configure node identity information, and start the cluster machine; Step 2: Each node generates a determined MPT-based data storage structure through a consensus mechanism, and enters a data submission stage after reaching a consensus on the structure; Step 3: In the data submission process, the encoding module in each replica node is triggered to collect the MPT nodes generated in the block execution process, and organizes and divides these nodes into multiple minimum data blocks; Step 4: The encoding module uses a multi-level hierarchical erasure code encoding method to encode the data blocks generated in step 3, wherein the bottom layer encoding is used to support lightweight incremental recovery, and the upper layer encoding is used to ensure data availability in the presence of Byzantine faults; Step 5: The encoded data blocks and their corresponding redundant check blocks are distributed and stored in a decentralized manner to multiple replica nodes, so that different replica nodes hold part of the data blocks or redundant check blocks, realizing distributed storage and reading of data blocks and redundant check blocks in a decentralized environment. 2.The method of claim 1, wherein, The cluster of step 1 is a collection of server nodes in the network that perform the same function; the node identity information includes: node role, node IP, node ID and public and private keys; the node IP is used to represent the IP address of the server node; the node ID is used to uniquely identify the server node; the public and private keys are used to encrypt messages and authenticate identity. 3.The method of claim 1, wherein, The node organization division of step 3 divides the MPT tree into multiple independent tree node sets, and appends index information at the pruning node to support complete indexing and positioning of the tree node set.

4. The layered erasure code storage method for MPT partitioned scenarios according to claim 1, characterized in that, The multi-level hierarchical erasure code encoding method of step 4 uses an erasure code encoding method based on Reed-Solomon encoding.

5. The layered erasure code storage method for MPT partitioned scenarios according to claim 1, characterized in that, The decentralized distribution of step 5 is that each replica node negotiates the encoded data blocks and redundant check blocks through a deterministic protocol, wherein the allocation rules for allocating data blocks and redundant check blocks are agreed in advance, and the data content that each replica node needs to store is determined, realizing distributed storage of data blocks and redundant check blocks.

6. The layered erasure code storage method for MPT partitioned scenarios according to claim 1 or claim 4, characterized in that, The hierarchical erasure code encoding encodes different levels according to the granularity of node subsets, and different levels use different encoding parameters, wherein the bottom layer encoding uses lower redundancy erasure code parameters to support local and lightweight recovery, and the top layer encoding uses higher redundancy erasure code parameters to provide final data recovery guarantee in the case where the bottom layer encoding cannot recover.

7. The layered erasure code storage method for MPT partitioned scenarios according to claim 6, characterized in that, The node subsets are subjected to data blocking processing by the post-partition encoding module, specifically, multiple tree node data belonging to the same node subset, including the data of each tree node itself and the index information of the child nodes possibly contained by the tree node, are sequentially spliced according to a predetermined deterministic order to form at least one data block, and the data block is further divided into sub-blocks to serve as the minimum unit for erasure code decoding.

8. The layered erasure code storage method for MPT partitioned scenarios according to claim 6 or claim 7, characterized in that, The redundancy blocks generated by the erasure code encoding of the node subsets are organized as vector elements in a predetermined deterministic order, and the vector elements are committed to ensure the integrity and verifiability of the node subsets and the redundancy blocks.

9. The layered erasure code storage method for MPT partitioned scenarios according to claim 1, wherein, The decoding module is triggered to perform erasure code decoding operation to recover the required MPT node data when a state reading request occurs, the required MPT node data is detected to be unavailable, or the MPT node data fails to pass the integrity or consistency verification.

Images