Topology Modeling and Self-Recovery Method for Multi-Protocol Fusion Acquisition Terminals in Data Centers

Abstract

This invention discloses a method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers, relating to the field of intelligent operation and maintenance technology for data center communication networks. The method bypasses and monitors the communication data streams of heterogeneous devices, uses a syntax tree to identify protocols and calculate semantic conversion overhead, and constructs a cross-protocol compatibility matrix; it collects the temperature change rate and power load fluctuation characteristics of rack nodes, extracts thermoelectric coupling correlation to establish physical proximity mapping, reconstructs and generates a heterogeneous semantic routing graph, and marks hidden parallel links; it monitors the access of heterogeneous links, generates a channel bottleneck vector based on the ratio of channel throughput to protocol encapsulation depth, and updates the routing node weights; when the main link is interrupted, it performs field pruning on the communication data according to the channel bottleneck constraint, extracts state telemetry metadata, downgrades and reconstructs it into a low-bit-width protocol frame, and then sends it along the minimum weight path of the heterogeneous routing graph and hidden links.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for data center communication networks, specifically to a method for topology modeling of multi-protocol fusion acquisition terminals and self-recovery of heterogeneous links for data centers. Background Technology

[0002] Data centers, as the infrastructure supporting massive computing and storage tasks, house a large number of IT servers and heterogeneous sensing devices responsible for environmental monitoring. To achieve real-time monitoring of the entire data center's status, multi-protocol fusion acquisition terminals are typically used to aggregate the operating parameters of the underlying physical devices. With the expansion of business scale and the continuous iteration of hardware architecture, the communication network at the bottom of the data center exhibits a high degree of heterogeneity. Different types of physical devices access the acquisition network through diverse communication links and protocol specifications, resulting in dynamic changes in the connectivity and physical distribution of the underlying terminals.

[0003] In the process of topology modeling and fault self-healing in heterogeneous acquisition networks, existing topology discovery methods mainly rely on the active detection mechanism of homogeneous network layer protocols. This traversal method based on a single protocol stack is prone to creating data silos due to protocol parsing barriers when facing environments with multiple protocols coexisting at the underlying level. It judges physically adjacent but protocol-incompatible nodes as completely isolated, resulting in a serious disconnect between the topology mapping results and the actual physical association. At the same time, existing network fault recovery schemes usually adopt symmetrical bandwidth redundancy and homogeneous protocol switching strategies. When the main acquisition link is interrupted and the backup channel is a low-speed heterogeneous link, traditional methods often directly transmit the complete original protocol message. This hard switching mechanism, which does not distinguish between load levels, will instantly exceed the throughput limit of the low-speed channel, causing a large backlog of messages and network congestion. This results in the inability to transmit the underlying state telemetry data in a timely manner, which seriously affects the disaster recovery and convergence efficiency of the data center's underlying network under link failure. Therefore, it is still necessary to provide a topology modeling and heterogeneous link self-healing method for multi-protocol fusion acquisition terminals in data centers to improve the topology mapping accuracy and link failure fault tolerance and self-healing capability in heterogeneous environments. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides the following technical solution: a method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers, comprising the following steps: S1, bypassing and monitoring the communication data streams of heterogeneous devices in the data center, extracting the packet header structure and payload data fields, inputting them into a multi-protocol syntax tree for protocol type identification, calculating the semantic conversion overhead between protocols, and constructing a cross-protocol compatibility matrix; S2, collecting the temperature change rate and power load fluctuation value of the data center rack nodes, calculating the thermoelectric coupling correlation degree of the temperature change rate and power load fluctuation value, extracting nodes with non-zero thermoelectric coupling correlation degree to establish physical proximity mapping, reconstructing the physical proximity mapping and cross-protocol compatibility matrix into a heterogeneous semantic routing graph, and in the heterogeneous semantic routing graph... S3. Mark nodes with disconnected communication and physical proximity mapping as hidden parallel links; S4. Monitor the access status of heterogeneous links, analyze the protocol encapsulation depth and channel throughput of heterogeneous links, calculate the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector, map the channel bottleneck vector to the heterogeneous semantic routing graph and update the node connection weight value; S5. When the main acquisition link is interrupted, obtain the maximum payload byte number corresponding to the channel bottleneck vector, perform field pruning on the communication data stream according to the maximum payload byte number to extract status telemetry metadata, combine the cross-protocol compatibility matrix to downgrade and reconstruct the status telemetry metadata into a low-bit-width protocol frame, and send the low-bit-width protocol frame along the path with the minimum node connection weight value in the heterogeneous semantic routing graph and hidden parallel links.

[0006] In a preferred embodiment, the specific process of bypassing the monitoring of communication data streams of heterogeneous devices in the data center, extracting the message header structure and payload data fields, and inputting them into a multi-protocol syntax tree for protocol type identification is as follows: A network communication mirror port is established for the heterogeneous devices to capture the original bit stream in the physical link. A fixed offset interval in the original bit stream is extracted as the message header structure, and the remaining data segments are extracted as the payload data fields. The root node of the multi-protocol syntax tree is traversed, and the feature identifier bits in the message header structure are compared with the transition trigger conditions of the root node. When the feature identifier bits meet the transition trigger conditions, a depth-first traversal is performed along the multi-protocol syntax tree to the child nodes to parse the encoding rules of the payload data fields, and the protocol type label corresponding to the leaf node at the end of the multi-protocol syntax tree is extracted as the protocol type identification result.

[0007] In a preferred embodiment, the specific process of calculating the semantic conversion overhead between protocols and constructing a cross-protocol compatibility matrix is ​​as follows: extract the source protocol data dictionary and the target protocol data dictionary corresponding to the protocol type identification result, and compare the field mapping overlap between the source protocol data dictionary and the target protocol data dictionary; count the difference in the number of handshake interactions between the source protocol state machine and the target protocol state machine during the connection establishment process, and combine the difference in the number of handshake interactions with the field mapping overlap to map it into semantic conversion overhead; extract the node identifier of heterogeneous devices as the matrix dimension, and fill in the semantic conversion overhead in the matrix element positions corresponding to different node identifiers to generate a cross-protocol compatibility matrix.

[0008] In a preferred embodiment, the specific process of collecting the temperature change rate and power load fluctuation value of data center rack nodes, calculating the thermoelectric coupling correlation degree between the temperature change rate and power load fluctuation value, and extracting nodes with non-zero thermoelectric coupling correlation degree to establish physical proximity mapping is as follows: The data center environment sensing interface is called to obtain the temperature change rate and power load fluctuation value of rack nodes within a continuous time window; cross-correlation analysis is performed on the temperature change rate and power load fluctuation value in the time dimension, and the peak covariance output of the cross-correlation analysis is extracted as the thermoelectric coupling correlation degree; rack nodes with non-zero thermoelectric coupling correlation degree are clustered into co-located rack node clusters, and physical proximity mapping is established between rack nodes within each co-located rack node cluster.

[0009] In a preferred embodiment, the specific process of reconstructing the physical proximity mapping and cross-protocol compatibility matrix into a heterogeneous semantic routing graph, and marking nodes with communication disconnection and physical proximity mapping as hidden parallel links in the heterogeneous semantic routing graph, is as follows: Extract the node identifier set and semantic conversion overhead from the cross-protocol compatibility matrix; instantiate the node identifier set as network topology nodes in the graph model framework; instantiate the semantic conversion overhead as directed edge weights between network topology nodes to generate an initial semantic routing graph; traverse the co-location cabinet node clusters in the physical proximity mapping; retrieve disconnected node pairs with empty directed edge weights between corresponding nodes in the co-location cabinet node clusters in the initial semantic routing graph; insert virtual direct edges between disconnected node pairs and assign near-field communication interface identifiers; update the initial semantic routing graph after inserting virtual direct edges into a heterogeneous semantic routing graph; and mark the virtual direct edges as hidden parallel links.

[0010] In a preferred embodiment, the specific process of monitoring the access status of the heterogeneous link, parsing the protocol encapsulation depth and channel throughput, and calculating the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector is as follows: Monitoring the level transition events of the physical layer of the communication interface to identify the access status of the heterogeneous link; intercepting the initial handshake message of the heterogeneous link, decapsulating the network layer and data link layer header tags of the initial handshake message layer by layer, and calculating the total length of the nested headers as the protocol encapsulation depth; reading the hardware configuration register of the network interface card corresponding to the heterogeneous link, and extracting the maximum transmission rate locked by the hardware configuration register as the channel throughput; mapping the channel throughput to an available bandwidth scalar, mapping the protocol encapsulation depth to a transmission overhead scalar, calculating the ratio of the available bandwidth scalar to the transmission overhead scalar, and constructing a channel bottleneck vector containing the ratio, the available bandwidth scalar, the transmission overhead scalar, and their constraint relationship.

[0011] In a preferred embodiment, the specific process of mapping the channel bottleneck vector to the heterogeneous semantic routing graph and updating the node connection weight values ​​is as follows: extract the available bandwidth scalar, transmission overhead scalar, and ratio from the channel bottleneck vector; locate the network topology node coordinates corresponding to the heterogeneous link in the heterogeneous semantic routing graph; create new directed edges between the network topology node coordinates; combine the constraint relationship between the transmission overhead scalar and the available bandwidth scalar with the ratio to transform it into a network congestion penalty factor; perform regularization fusion of the network congestion penalty factor and the original node connection weight values ​​in the heterogeneous semantic routing graph; and update the node connection weight values ​​in the new directed edges and the heterogeneous semantic routing graph.

[0012] In a preferred embodiment, when the main acquisition link is interrupted, the specific process of obtaining the maximum payload byte count corresponding to the channel bottleneck vector and performing field pruning on the communication data stream to extract status telemetry metadata based on the maximum payload byte count is as follows: Configure a heartbeat keep-alive timer, and determine that the main acquisition link is interrupted when the heartbeat keep-alive timer times out and no response frame is received from the main acquisition link; parse the constraint relationship contained in the channel bottleneck vector, and retrieve the physical channel maximum transmission unit threshold supported by the constraint relationship as the maximum payload byte count; perform reverse parsing on the message structure of the communication data stream according to the maximum payload byte count, and remove the regular system logs and descriptive text in the message structure as a field pruning step, and extract the status telemetry metadata containing device operating parameters and fault codes in the message structure.

[0013] In a preferred embodiment, the specific process of downgrading and reconstructing state telemetry metadata into low-width protocol frames by combining a cross-protocol compatibility matrix, and sending low-width protocol frames along the path with the minimum node connection weight value in the heterogeneous semantic routing graph and hidden parallel links is as follows: Query the cross-protocol compatibility matrix, match the source protocol identifier of the state telemetry metadata with the target protocol identifier in the cross-protocol compatibility matrix, and select the target low-width protocol template with the minimum semantic conversion overhead; encapsulate the state telemetry metadata into the payload range of the target low-width protocol template, recalculate the communication checksum to complete the downgrading and reconstruction to generate low-width protocol frames; in the global topology space of the heterogeneous semantic routing graph and hidden parallel links, use the node connection weight value as the pathfinding heuristic cost to perform shortest path retrieval to generate a recovery route sequence, and forward the low-width protocol frames along the recovery route sequence.

[0014] The technical effects and advantages of this invention are as follows:

[0015] (1) A method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers is proposed. In the early modeling stage, protocol type identification and semantic conversion overhead are performed by extracting communication message features to generate a cross-protocol compatibility matrix. A physical proximity mapping is established by combining the thermoelectric coupling correlation between the rack node operating temperature change rate and the power supply load fluctuation value. Then, the physical proximity mapping and the cross-protocol compatibility matrix are reconstructed into a heterogeneous semantic routing graph to discover hidden parallel links that are physically adjacent but have disconnected communication. As a result, the network island phenomenon caused by the existing single-protocol detection can be overcome. The spatiotemporal correlation characteristics of the physical environment are introduced into the topology modeling process to correct the mapping error of the logical network and the physical space being disconnected, and to increase the topology connectivity dimension and the number of potential backup channels of the underlying acquisition network.

[0016] (2) A multi-protocol fusion acquisition terminal topology modeling and heterogeneous link self-recovery method for data centers is proposed. During network operation, the protocol encapsulation depth and physical channel throughput of the heterogeneous links are analyzed to generate channel bottleneck vectors and update routing node weights. When the main acquisition link is interrupted, the original data stream is pruned according to the maximum payload byte number constrained by the channel bottleneck vector to extract core state telemetry metadata. Combined with the cross-protocol compatibility matrix, it is downgraded and reconstructed into low-bit-width protocol frames and sent along the hidden parallel link. This solves the message backlog and congestion problems caused by the existing homogeneous switching strategy in low-speed heterogeneous backup channels. Through adaptive semantic pruning and protocol format dimensionality reduction of communication data, it ensures that the critical state data matches the throughput limit of the asymmetric channel in extreme link failure scenarios, and guarantees the cross-media fault-tolerant transmission efficiency of the underlying monitoring data of the data center.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] Figure 1 This is a flowchart of the multi-protocol fusion acquisition terminal topology modeling and heterogeneous link self-recovery method for data centers according to the present invention.

[0019] Figure 2 This is a time series analysis diagram illustrating the thermoelectric coupling spatiotemporal correlation between the temperature change rate and the power supply load fluctuation value in an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the capacity constraint mechanism for performing field pruning on the communication data stream based on the maximum payload byte count in an embodiment of the invention. Detailed Implementation

[0021] This application's embodiments address the problems of existing heterogeneous acquisition network topology modeling being disconnected from the physical environment and heterogeneous backup link congestion failure when the backbone link is disconnected by using a multi-protocol fusion acquisition terminal topology modeling and heterogeneous link self-recovery method for data centers.

[0022] The overall approach of the scheme in this application is as follows: First, the heterogeneous communication data intercepted by the bypass is parsed and the semantic conversion overhead is evaluated. At the same time, the synchronous fluctuation characteristics of the thermal and power loads of the data center rack equipment are collected. The two are combined to deduce the physical spatial adjacency relationship and construct a global heterogeneous semantic routing graph that includes conventional network connections and hidden parallel links. Then, the physical bandwidth and protocol encapsulation of the heterogeneous links are monitored in real time to quantify the channel bottleneck and correct the network routing weight. After confirming that the main acquisition link is interrupted, the original communication message is structurally pruned using the load limit limited by the channel bottleneck. Redundant fields are removed and key telemetry data is retained. The message is reassembled into a low-bit-width format frame according to the protocol conversion overhead and forwarded along the pre-discovered optimal hidden parallel link.

[0023] Example 1; please refer to Figure 1This invention provides a technical solution: a method for multi-protocol fusion acquisition terminal topology modeling and heterogeneous link self-recovery for data centers, comprising the following steps: S1, bypassing and monitoring the communication data streams of heterogeneous devices in the data center, extracting the packet header structure and payload data fields, inputting them into a multi-protocol syntax tree for protocol type identification, calculating the semantic conversion overhead between protocols, and constructing a cross-protocol compatibility matrix; S2, collecting the temperature change rate and power load fluctuation value of the data center rack nodes, calculating the thermoelectric coupling correlation degree of the temperature change rate and power load fluctuation value, extracting nodes with non-zero thermoelectric coupling correlation degree to establish physical proximity mapping, reconstructing the physical proximity mapping and cross-protocol compatibility matrix into a heterogeneous semantic routing graph, and integrating the communication data streams of heterogeneous devices in the heterogeneous semantic routing graph into a heterogeneous semantic routing graph. Nodes with disconnected communication and physical proximity mapping are marked as hidden parallel links; S3, monitor the access status of heterogeneous links, analyze the protocol encapsulation depth and channel throughput of heterogeneous links, calculate the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector, map the channel bottleneck vector to the heterogeneous semantic routing graph and update the node connection weight value; S4, when the main acquisition link is interrupted, obtain the maximum number of payload bytes corresponding to the channel bottleneck vector, perform field pruning on the communication data stream based on the maximum number of payload bytes to extract status telemetry metadata, combine the cross-protocol compatibility matrix to downgrade and reconstruct the status telemetry metadata into a low-width protocol frame, and send the low-width protocol frame along the path with the minimum node connection weight value in the heterogeneous semantic routing graph and hidden parallel links.

[0024] In this implementation scheme, step S1 primarily involves replicating and analyzing network-side traffic to identify the protocol types of different devices and quantify the difficulty of data conversion between different protocols. During this process, bypass monitoring refers to a method of mirroring and copying network data streams for analysis without interrupting the original communication link. A multi-protocol syntax tree (MPLT) is a hierarchical data structure used to disassemble message header features layer by layer according to protocol standards to match corresponding protocol specifications. Semantic conversion overhead characterizes the processor resources and time latency consumed when parsing and re-encapsulating data packets of one protocol into another. The cross-protocol compatibility matrix is ​​a two-dimensional table recording the cost values ​​for mutual conversion between all identified protocols within the data center, providing basic data for subsequent cross-protocol routing.

[0025] Step S2 analyzes the synchronization patterns of heat generation and power consumption of equipment within the computer room to deduce the physical adjacency relationships of the equipment in the real three-dimensional space, thereby uncovering backup communication channels that cannot be detected by the network layer. Thermoelectric coupling correlation represents the statistical correlation between the rise and fall trends of equipment operating temperature and the frequency of power fluctuations; a higher correlation indicates that the equipment is highly likely to be located in the same physical cabinet or the same power supply branch. Physical proximity mapping is a table of spatial adjacency relationships of equipment established based on the above correlations. The heterogeneous semantic routing graph is a comprehensive network topology map that integrates conventional network connections, protocol conversion costs, and physical spatial adjacency relationships. Hidden parallel links refer to out-of-band backup channels that are isolated from each other in conventional Ethernet but are physically close, enabling temporary data transmission through near-field physical interfaces when the main network fails.

[0026] Step S3 is used to dynamically evaluate the actual transmission capacity of new links when adding new equipment during data center expansion, and update the routing parameters of the entire network accordingly. An additional link refers to a network connection added during data center network expansion that may differ from the backbone network in physical medium or protocol type. Protocol encapsulation depth refers to the total number of bytes of network layer and data link layer header tags nested and wrapped before data transmission; deeper nesting consumes more bandwidth. Channel throughput refers to the maximum physical transmission rate supported by the network interface card hardware. The channel bottleneck vector is an evaluation metric that reflects the net bandwidth capacity of the link to carry effective data after deducting protocol header overhead. By updating node connection weights, the system can prevent large data packets from being scheduled to congested links that appear to have high bandwidth but have extremely high protocol overhead.

[0027] Step S4, when a network outage occurs in the main communication network, performs data payload dimensionality reduction and format conversion to ensure that the most critical alarm information can be transmitted through the low-speed backup channel. Maximum payload bytes refer to the maximum effective data capacity that the backup channel can transmit in a single transmission without packet fragmentation. Field pruning refers to the process of reversely disassembling the original data packet and deleting unnecessary information such as routine operation logs. Status telemetry metadata refers to the critical data retained only after pruning, such as equipment fault codes and core operating parameters. Downgrading and reconstructing to low-bandwidth protocol frames refers to converting the high-bandwidth network protocol format into a concise and efficient basic industrial control protocol format. Finally, based on the calculated weight values, the system relays the slimmed-down core data through physically adjacent backup channels, completing link self-recovery under fault conditions.

[0028] Specifically, the process of bypassing and monitoring the communication data streams of heterogeneous devices in the data center, extracting the message header structure and payload data fields, and inputting them into a multi-protocol syntax tree for protocol type identification is as follows: A network communication mirror port is established for the heterogeneous devices to capture the original bit stream in the physical link. A fixed offset interval in the original bit stream is extracted as the message header structure, and the remaining data segments are extracted as the payload data fields. The root node of the multi-protocol syntax tree is traversed, and the feature identifier bits in the message header structure are compared with the transition trigger conditions of the root node. When the feature identifier bits meet the transition trigger conditions, a depth-first traversal is performed along the multi-protocol syntax tree to the child nodes to parse the encoding rules of the payload data fields, and the protocol type label corresponding to the leaf node at the end of the multi-protocol syntax tree is extracted as the protocol type identification result.

[0029] In this implementation, the high-level and low-level signals on the physical link side are converted into a raw bitstream containing a complete frame structure via the network communication mirror port. The bitstream is then segmented according to a preset fixed offset cursor, separating independent header feature bit sets and payload data segments. The system inputs the header feature bit set into the root node of the multi-protocol syntax tree for matching. When the feature bits meet the transition trigger condition of the root node, a downward depth-first traversal process is initiated. Upon reaching each child node, the structural fit between the payload data segment and the preset encoding rules of the current child node is calculated. The specific calculation formula is as follows: ;in, This indicates the structural fit between the payload data segment and the current child node encoding rule; Indicates the total bit length of the payload data segment; This indicates the total number of independent data blocks into which the payload data segment is divided; m represents the index number of the independent data block. This represents the positional weight coefficient corresponding to the m-th independent data block; This represents the actual bit content of the m-th independent data block; This represents the m-th pre-defined standard parsing template of the current syntax tree child node c; This represents a bitwise XOR matching function between the actual bit content and the standard parsing template; This represents the format penalty factor for the m-th independent data block; This represents the percentage of non-canonical padding characters in the actual bit content. Position weight coefficient. The determination method involves statistically analyzing the frequency of status control commands contained in the data block at that location in the device's historical communication records, normalizing the frequency according to its proportion of the total communication frequency, and then assigning a value. The structure fit is calculated by continuously traversing downwards along the syntax tree until the final leaf node is reached and all fit values ​​are greater than a preset benchmark. The protocol type tag bound to that leaf node is then extracted to complete the final identification.

[0030] Specifically, the process of calculating the semantic conversion overhead between protocols and constructing a cross-protocol compatibility matrix is ​​as follows: Extract the source protocol data dictionary and the target protocol data dictionary corresponding to the protocol type identification results, and compare the overlap of field mappings between the source protocol data dictionary and the target protocol data dictionary; Calculate the difference in the number of handshake interactions between the source protocol state machine and the target protocol state machine during the connection establishment process, and combine the difference in the number of handshake interactions with the overlap of field mappings to map the semantic conversion overhead; Extract the node identifiers of heterogeneous devices as the matrix dimension, and fill in the semantic conversion overhead in the matrix element positions corresponding to different node identifiers to generate a cross-protocol compatibility matrix.

[0031] In this implementation scheme, based on the identified protocol type label, the corresponding source protocol data dictionary and target protocol data dictionary are retrieved from the local repository. The mapping overlap between the two at the field definition level is analyzed. At the same time, the difference in the number of interactions between the source protocol state machine and the target protocol state machine during the connection handshake operation is tracked. The overlap and the difference in the number of interactions are jointly deduced and mapped into semantic transformation overhead. The specific calculation formula is as follows: ;in, This represents the semantic conversion overhead of converting source protocol a to target protocol b; Indicates the weight of the field mapping overlap constraint; This represents the total number of field categories contained in the union data dictionary of the source and target protocols; p represents the index number of the field category. This indicates the effective byte length defined by source protocol a under the p-th field category; This indicates the effective byte length defined by the target protocol b under the p-th field category; This indicates the length of the available mapping intersection between the two under this field category; This indicates the length of the union of the structures of the two objects under this field category; Indicates the constraint weight for the difference in the number of handshake interactions; This indicates the total number of handshake stages required to establish a communication connection; Indicates the index number of the handshake phase; Indicates source protocol In the The number of communication interactions consumed by the state transition in each handshake phase; This represents the number of communication interactions consumed by target protocol b during the state transition in the q-th handshake phase. Weight parameter. and The method for determining this is to monitor the service distribution characteristics of the underlying topology of the data center, and increase the speed when the proportion of latency-sensitive devices exceeds half. Increase the load when more than half of the nodes are sensitive to packet loss. Furthermore, the sum of the two values ​​is forced to be equal to 1. After the calculation is completed, a two-dimensional matrix is ​​constructed using the physical node identifiers of the heterogeneous devices connected to the acquisition terminal as the row and column dimensions. The corresponding semantic conversion overhead is filled into the coordinate positions of the elements where different node identifiers intersect, forming a complete cross-protocol compatibility matrix.

[0032] Specifically, the process of collecting the temperature change rate and power load fluctuation value of data center rack nodes, calculating the thermoelectric coupling correlation degree between the temperature change rate and power load fluctuation value, and extracting nodes with non-zero thermoelectric coupling correlation degree to establish physical proximity mapping is as follows: The data center environment sensor interface is called to obtain the temperature change rate and power load fluctuation value of rack nodes within a continuous time window; cross-correlation analysis is performed on the temperature change rate and power load fluctuation value in the time dimension, and the peak covariance output of the cross-correlation analysis is extracted as the thermoelectric coupling correlation degree; rack nodes with non-zero thermoelectric coupling correlation degree are clustered into co-located rack node clusters, and physical proximity mapping is established between rack nodes within each co-located rack node cluster.

[0033] In this implementation scheme, the system uses an environmental monitoring network deployed within the data center to call environmental sensor interfaces at a fixed sampling frequency to acquire real-time temperature and power supply data for each rack node within a continuous sliding window period. The system then calculates the gradient of these data over time to obtain time series data for temperature change rate and power load fluctuation. Since the power surge of servers within the same physical rack during computation can instantly trigger power load fluctuations, and the dissipated heat can cause the local ambient temperature to rise synchronously with a very short lag, a cross-correlation analysis considering time offset is performed on these two heterogeneous time series. The peak covariance in the analysis results is extracted to quantify this thermoelectric coupling effect. The calculation formula is as follows: ;in, Indicates a time delay of At time, the cross-correlation coefficient between the temperature change rate sequence of node x and the power supply load fluctuation sequence of node y; W represents the total number of sampling points included in the continuous time window; t represents the discrete time sampling points; This represents the actual rate of temperature change of rack node x at time t; This represents the statistical average of the rate of temperature change at node x within a continuous time window. This indicates that rack node y has a delay time of... The moment The actual power supply load fluctuation value; This represents the statistical average value of the power supply load fluctuation at node y within a continuous time window. This represents the thermoelectric coupling degree between node x and node y; This represents the maximum allowable lag time threshold. Maximum Lag Time Threshold The method for determining this threshold involves retrieving the airflow velocity of the air conditioning system in the hot and cold aisles of the computer room and the spacing between server racks. The longest propagation time required for cold air to pass through two adjacent server racks is used as the upper limit of this threshold. After calculation, the system removes the thermoelectric coupling correlation. A pair of nodes equal to zero indicates that they do not affect each other in physical space; Rack nodes with values ​​significantly greater than zero are grouped into clusters based on hierarchical clustering algorithms. Strongly related nodes are grouped into clusters within the same rack, and these clustered nodes are then labeled with physical adjacency tags within the system database. For example... Figure 2 The figure shows a time series analysis diagram of the thermoelectric coupling spatiotemporal correlation between the temperature change rate and the power supply load fluctuation value in an embodiment of the present invention. The horizontal axis represents discrete time sampling points (t1 to t9), the left vertical axis represents the power supply load fluctuation value, and the right vertical axis represents the temperature change rate. As can be seen from the figure, the power supply load fluctuation value curve (solid line) and the temperature change rate curve (dashed line) of the rack node within a continuous time window exhibit a high degree of cross-correlation. When the server's computing load increases sharply, generating a peak, the heat it dissipates causes a local temperature increase, resulting in a slight "lag time" in the peak of the temperature change rate curve relative to the peak of the power supply load fluctuation value on the time axis. The system quantifies the thermoelectric coupling correlation by performing a cross-correlation analysis considering the time offset on these two heterogeneous time series and extracting their peak covariance.

[0034] Specifically, the process of reconstructing the physical proximity mapping and cross-protocol compatibility matrix into a heterogeneous semantic routing graph, and marking nodes with communication disconnection and physical proximity mapping as hidden parallel links in the heterogeneous semantic routing graph, is as follows: Extract the node identifier set and semantic transformation overhead from the cross-protocol compatibility matrix; instantiate the node identifier set as network topology nodes in the graph model framework; instantiate the semantic transformation overhead as directed edge weights between network topology nodes to generate an initial semantic routing graph; traverse the co-location cabinet node clusters in the physical proximity mapping; retrieve disconnected node pairs with empty directed edge weights between corresponding nodes in the co-location cabinet node clusters in the initial semantic routing graph; insert virtual direct edges between disconnected node pairs and assign near-field communication interface identifiers; update the initial semantic routing graph after inserting virtual direct edges to a heterogeneous semantic routing graph; and mark the virtual direct edges as hidden parallel links.

[0035] In this implementation, the system first reads the cross-protocol compatibility matrix constructed in the previous stage and transforms it into a directed weighted graph data structure in graph theory. Within this graph model framework, the node identifiers of various heterogeneous devices represented by the rows and columns of the matrix are directly instantiated as sets of entity nodes in the network topology graph. The non-empty elements in the matrix, representing the semantic conversion overhead between protocols, are instantiated as the initial weights of the directed edges between the corresponding nodes, thereby generating an initial semantic routing graph reflecting the current network protocol connectivity status. Subsequently, the system performs a dual search in the initial semantic routing graph, using the clusters of nodes in the same rack as defined in the physical proximity mapping. On one hand, it confirms whether these nodes belong to the same rack cluster; on the other hand, it checks whether the weights of their directed edges in the routing graph model are null. If it is found that node A and node B belong to the same rack cluster in the physical mapping (are very close), but the edge weights in the routing graph are null (meaning that the protocol is not working at the conventional Ethernet or master network level, or the physical network cable is disconnected), then such a node pair is defined as a disconnected node pair. For the identified disconnected node pairs, the system forcibly inserts a virtual direct connection edge connecting the two in the network topology graph, and forcibly assigns the interface type of this direct connection edge to a near-field communication interface with short-range communication capabilities, such as Bluetooth, infrared, or a dedicated serial port. After the above insertion and labeling operations, the topology graph, which integrates the real network connectivity and physical space connectivity potential, is officially updated into a heterogeneous semantic routing graph. These newly inserted virtual direct connections edges are uniformly marked as hidden parallel links by the global system. They are normally in a dormant state, but can serve as backup transmission channels for short-range relay transmission of alarm data when the backbone network fails.

[0036] Specifically, the process of monitoring the access status of the heterogeneous link, parsing the protocol encapsulation depth and channel throughput, and calculating the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector is as follows: Monitoring the level transition events of the physical layer of the communication interface to identify the access status of the heterogeneous link; intercepting the initial handshake message of the heterogeneous link, decapsulating the network layer and data link layer header tags of the initial handshake message layer by layer, and calculating the total length of the nested headers as the protocol encapsulation depth; reading the hardware configuration register of the network interface card corresponding to the heterogeneous link, extracting the maximum transmission rate locked by the hardware configuration register as the channel throughput; mapping the channel throughput to an available bandwidth scalar, mapping the protocol encapsulation depth to a transmission overhead scalar, calculating the ratio of the available bandwidth scalar to the transmission overhead scalar, and constructing a channel bottleneck vector containing the ratio, the available bandwidth scalar, the transmission overhead scalar, and their constraint relationship.

[0037] In this implementation, the system's underlying driver continuously polls the physical layer transceiver status register of the communication interface. When it detects a sudden change in pin level from a continuously low idle state to a high-frequency toggle active state, it triggers a hardware interrupt to identify the physical access status of the heterogeneous link. Subsequently, the system's underlying network processor intercepts the initial handshake message sent by this link to establish a connection. Following the protocol stack order, it decapsulates the header tags of the data link layer and network layer layer by layer from bottom to top, extracts the actual space occupied by the stripped tags at each layer, and accumulates them. The total length of the nested header obtained by summing the accumulated data is defined as the protocol encapsulation depth. Simultaneously, the system sends a configuration space read command to the network interface card corresponding to the heterogeneous link via the internal bus to obtain the maximum physical transmission rate stored in the hardware register as the channel throughput. After this physical parameter extraction is complete, scalar conversion and ratio calculation are performed to generate the channel bottleneck vector. The specific calculation formula is as follows: ;in, This represents the core ratio upon which the generation of the channel bottleneck vector depends; This represents the available bandwidth scalar generated by directly mapping the channel throughput; Indicates the number of nested header tags that were successfully parsed; k represents the index sequence number of the header tag nesting level; This indicates the length of bytes occupied by the header tag of the k-th protocol layer; This represents the transmission overhead scalar generated by the protocol encapsulation depth mapping; This represents the physical inter-frame gap compensation constant to prevent division-by-zero anomalies. This represents the final constructed channel bottleneck vector; This represents a nonlinear throughput constraint function that reflects the interdependence between the available bandwidth scalar and the transmission overhead scalar. Physical inter-frame compensation constant. The method for determining this is to read the equivalent value of the forced silence period bytes that are fixed in the minimum frame transmission mode of the current network interface card.

[0038] Specifically, the process of mapping the channel bottleneck vector to the heterogeneous semantic routing graph and updating the node connection weights is as follows: Extract the available bandwidth scalar, transmission overhead scalar, and ratio from the channel bottleneck vector; locate the network topology node coordinates corresponding to the heterogeneous links in the heterogeneous semantic routing graph; create new directed edges between the network topology node coordinates; combine the constraint relationship between the transmission overhead scalar and the available bandwidth scalar with the ratio to transform it into a network congestion penalty factor; perform regularization fusion of the network congestion penalty factor and the original node connection weights in the heterogeneous semantic routing graph; and update the node connection weights in the new directed edges and the heterogeneous semantic routing graph.

[0039] In this implementation scheme, the system first parses the channel bottleneck vector generated in the previous step, extracting the internally encapsulated available bandwidth scalar, transmission overhead scalar, and ratio. Then, in the global topology coordinate system of the heterogeneous semantic routing graph, based on the physical network interface card addresses of the devices at both ends of the heterogeneous link, the system precisely locates the coordinates of the corresponding network topology nodes and instantiates a new directed edge representing physical connectivity between the coordinates. To transform the asymmetric transmission resistance of heterogeneous channels into a routing addressing basis, the system uses a nonlinear transformation to fuse the scalar constraint relationship and the ratio into a network congestion penalty factor, followed by a global weight update and replacement. The specific calculation formula is as follows: ;in, This represents the network congestion penalty factor generated by the heterogeneous link mapping; express Extracted transmission overhead scalar; This represents the available bandwidth scalar value extracted. This represents the extracted ratio parameter; This represents the sensitivity coefficient to bandwidth overhead constraints. This indicates a non-linear adjustment weight for the ratio; This indicates that a smoothing constant is added to the denominator to prevent overflow; This represents the connection weight value of the updated node in the heterogeneous semantic routing graph; This represents the original node connection weight value in the heterogeneous semantic routing graph. For the newly established directed edges, its initial default value is the connectivity constant defined by the system. Indicates the historical topological inertia preservation factor; and These represent the minimum and maximum historical network congestion penalty factors recorded in the current global heterogeneous semantic routing graph buffer pool, respectively. Historical topology inertia retention factor. The determination method is based on the frequency of changes to the underlying topology within the past calendar month in the statistics center. The higher the frequency of changes, the more likely it is to be determined. The smaller the value, the faster the convergence and global effect of the newly added link weights.

[0040] Specifically, when the main acquisition link is interrupted, the process of obtaining the maximum payload byte count corresponding to the channel bottleneck vector and extracting status telemetry metadata by performing field pruning on the communication data stream based on the maximum payload byte count is as follows: Configure a heartbeat keep-alive timer, and determine that the main acquisition link is interrupted when the heartbeat keep-alive timer times out and no response frame is received from the main acquisition link; parse the constraint relationship contained in the channel bottleneck vector, and retrieve the physical channel maximum transmission unit threshold supported by the constraint relationship as the maximum payload byte count; perform reverse parsing on the message structure of the communication data stream based on the maximum payload byte count, and remove the regular system logs and descriptive text in the message structure as a field pruning step, and extract the status telemetry metadata containing device operating parameters and fault codes in the message structure.

[0041] In this implementation scheme, the system configures an independently operating heartbeat keep-alive timer on the underlying network interface to periodically send probes to the peer device. If no corresponding acknowledgment frame is captured within several consecutive timing periods after the probe is sent, the system hardware interrupt controller sets the connectivity flag of the current main acquisition link to failure, formally determining that a link interruption has occurred. Subsequently, the system retrieves the channel bottleneck vector generated in the previous steps, parses the internal throughput constraint relationship function, and extracts the critical tolerance value that the physical channel will not experience physical layer slice packet loss under the current transmission overhead constraints, using this as the maximum payload bytes. To ensure that core data is successfully transmitted within this payload limit, the system performs reverse syntax tree parsing on the raw communication data stream remaining in the transmission queue, cutting it into multiple independent field blocks. As a field pruning step, blocks marked as regular system logs and descriptive text are directly discarded. For the remaining candidate field blocks, the system constructs a constraint optimization-based model to extract state telemetry metadata. The specific extraction formula is as follows: The extraction formula must satisfy the following capacity constraints: ;in, This represents the set of state telemetry metadata extracted after solving; This indicates the total number of candidate field blocks remaining after the original message is reverse-parsed and segmented; g represents the sequence index number of the candidate field block. This represents the criticality score of the g-th candidate field block; This represents a binary decision variable, where a value of 1 indicates that the field block is retained, and a value of 0 indicates that the field block is pruned or discarded. This indicates the actual byte length occupied by the g-th candidate field block; This represents the threshold for the maximum number of payload bytes retrieved from the channel bottleneck vector. Keyness score. The method for determining the weight is to scan the byte sequence of the field block using a feature matching engine. If it contains a preset device hardware fault code, it is assigned the highest weight score; if it only contains normal operating status parameters such as fan speed or voltage ripple, it is assigned a basic score. By solving this constraint equation, the system completes the compression of message semantics and the lossless extraction of core telemetry data while strictly adhering to the physical carrying capacity limits of the backup channel. Figure 3The diagram illustrates the capacity constraint mechanism for field pruning of communication data streams based on the maximum payload byte count. The horizontal axis compares different message processing stages (original message structure vs. pruned message structure), the vertical axis represents the actual byte length of the data packet, and the horizontal dashed line marks the physical channel maximum payload byte count constraint threshold resolved from the channel bottleneck vector. As can be seen, the total byte length of the original message structure far exceeds this capacity threshold, easily causing congestion in low-speed heterogeneous channels. After the system performs reverse parsing, as a field pruning step, the regular system logs and descriptive text blocks (gray area) located above the stacked histogram are completely removed, leaving only the bottom state telemetry metadata blocks (black area). The actual byte length of the pruned message structure is strictly converged to below the dashed line representing the maximum payload byte count, ensuring that the underlying critical monitoring data can match the throughput limit of the asymmetric channel under extreme link failure scenarios.

[0042] Specifically, the process of downgrading and reconstructing state telemetry metadata into low-width protocol frames by combining a cross-protocol compatibility matrix, and sending low-width protocol frames along the path with the minimum node connection weight value in heterogeneous semantic routing graphs and hidden parallel links is as follows: Query the cross-protocol compatibility matrix, match the source protocol identifier of the state telemetry metadata with the target protocol identifier in the cross-protocol compatibility matrix, and select the target low-width protocol template with the minimum semantic conversion overhead; encapsulate the state telemetry metadata into the payload range of the target low-width protocol template, recalculate the communication checksum to complete the downgrading and reconstruction, and generate low-width protocol frames; within the global topology space of the heterogeneous semantic routing graph and hidden parallel links, use the node connection weight value as the pathfinding heuristic cost to perform shortest path retrieval to generate a recovery route sequence, and forward the low-width protocol frames along the recovery route sequence.

[0043] In this implementation scheme, after extracting the state telemetry metadata, the system retrieves the constructed cross-protocol compatibility matrix. Using the source protocol identifier of the current data as the row index of the matrix, it traverses all non-empty column elements in the row, locates the coordinates of the element with the smallest semantic conversion cost, and selects the protocol type corresponding to that coordinate as the target low-bit-width protocol template. The system re-encodes the state telemetry metadata according to the bit-width specification of the target template and fills it into its payload range. It then recalculates the cyclic redundancy checksum of the frame header and payload using polynomial division and appends it to the frame tail, completing the data format downgrading and reconstruction to generate a low-bit-width protocol frame. Subsequently, within the global topology space composed of the updated weighted heterogeneous semantic routing graph and hidden parallel links, the system initiates a heuristic shortest path retrieval algorithm. The specific pathfinding cost evaluation formula is as follows: ;in, This indicates traversing to intermediate nodes in the topological space. Comprehensive pathfinding assessment cost at that time; Indicates the starting device node for currently sending low-width protocol frames; This indicates the target endpoint node to which the data needs to be aggregated; This represents the distance from the starting point to the current intermediate node in the search tree. The set of connected edges that have been explored; e represents a single directed edge or hidden parallel link in the set of connected edges; This represents the latest node connection weight value corresponding to the e-th edge; Indicates the current intermediate node and the target endpoint node Minimum number of logical hops in a graph theory model; This represents the hop count penalty coefficient. The method for determining this is to monitor the current concurrent thread load rate of the microprocessor in the acquisition terminal in real time. When the processor load rate exceeds the baseline operating threshold, the load rate is increased proportionally in exponential increments. The numerical values ​​are used to encourage the algorithm to choose paths with fewer hops and relay nodes during computation. The system relays data along the link sequence with the lowest overall pathfinding cost, achieving communication recovery in extreme link loss environments.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0049] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers, characterized in that, Includes the following steps: S1. Bypass monitoring of communication data streams of heterogeneous devices in the data center, extract message header structure and payload data fields, input them into a multi-protocol syntax tree for protocol type identification, calculate semantic conversion overhead between protocols, and construct a cross-protocol compatibility matrix; S2. Collect the temperature change rate and power load fluctuation value of the data center rack nodes, calculate the thermoelectric coupling correlation between the temperature change rate and power load fluctuation value, extract nodes with non-zero thermoelectric coupling correlation to establish physical proximity mapping, reconstruct the physical proximity mapping and cross-protocol compatibility matrix into a heterogeneous semantic routing graph, and mark nodes with communication disconnection and physical proximity mapping as hidden parallel links in the heterogeneous semantic routing graph. S3. Monitor the access status of heterogeneous links, analyze the protocol encapsulation depth and channel throughput of heterogeneous links, calculate the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector, map the channel bottleneck vector to the heterogeneous semantic routing graph and update the node connection weight value. S4. When the main acquisition link is interrupted, obtain the maximum number of payload bytes corresponding to the channel bottleneck vector, perform field pruning on the communication data stream based on the maximum number of payload bytes to extract state telemetry metadata, and combine the cross-protocol compatibility matrix to downgrade and reconstruct the state telemetry metadata into a low-width protocol frame. Send the low-width protocol frame along the path with the minimum node connection weight value in the heterogeneous semantic routing graph and hidden parallel link.

2. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 1, characterized in that: The specific process of bypassing the monitoring of communication data streams from heterogeneous devices in a data center, extracting the message header structure and payload data fields, and inputting them into a multi-protocol syntax tree for protocol type identification is as follows: Establish a network communication mirror port for heterogeneous devices to capture the original bit stream in the physical link, extract a fixed offset range from the original bit stream as the message header structure, and extract the remaining data fragments as the payload data field. Traverse the root node of the multi-protocol syntax tree and compare the feature identifier bits in the message header structure with the transition triggering conditions of the root node. When the feature identifier meets the transition triggering condition, a depth-first traversal is performed along the multi-protocol syntax tree to the child nodes to parse the encoding rules of the payload data field, and the protocol type label corresponding to the leaf node at the end of the multi-protocol syntax tree is extracted as the protocol type identification result.

3. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 2, characterized in that: The specific process of calculating the semantic conversion overhead between protocols and constructing a cross-protocol compatibility matrix is ​​as follows: Extract the source protocol data dictionary and target protocol data dictionary corresponding to the protocol type identification results, and compare the overlap of field mapping between the source protocol data dictionary and the target protocol data dictionary; The difference in the number of handshake interactions between the source protocol state machine and the target protocol state machine during the connection establishment process is statistically analyzed, and the difference in the number of handshake interactions is combined with the overlap of field mapping to map into semantic transformation overhead. Extract the node identifiers of heterogeneous devices as the matrix dimension, and fill in the semantic transformation overhead at the matrix element positions corresponding to different node identifiers to generate a cross-protocol compatibility matrix.

4. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 1, characterized in that: The specific process of collecting the temperature change rate and power load fluctuation value of data center rack nodes, calculating the thermoelectric coupling correlation between the temperature change rate and power load fluctuation value, and extracting nodes with non-zero thermoelectric coupling correlation to establish physical proximity mapping is as follows: The temperature change rate and power load fluctuation value of the rack node within a continuous time window are obtained by calling the data center environment sensor interface. Perform a time-dimensional cross-correlation analysis on the temperature change rate and power supply load fluctuation value, and extract the peak covariance of the cross-correlation analysis output as the thermoelectric coupling correlation degree. Rack nodes with non-zero thermo-electric coupling correlation are clustered into co-located cabinet node clusters, and physical proximity mappings are established between rack nodes within each co-located cabinet node cluster.

5. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 4, characterized in that: The physical proximity mapping and cross-protocol compatibility matrix are reconstructed into a heterogeneous semantic routing graph. The specific process of marking nodes with disconnected communication but physical proximity mapping as hidden parallel links in the heterogeneous semantic routing graph is as follows: Extract the node identifier set and semantic transformation overhead from the cross-protocol compatibility matrix. In the graph model framework, instantiate the node identifier set as network topology nodes and instantiate the semantic transformation overhead as directed edge weights between network topology nodes to generate an initial semantic routing graph. Traverse the clusters of co-located cabinet nodes in the physical proximity mapping and retrieve the disconnected node pairs with empty directed edge weights between corresponding nodes in the initial semantic routing graph. Insert virtual direct edges between disconnected node pairs and assign them near-field communication interface identifiers. Update the initial semantic routing graph after inserting virtual direct edges to a heterogeneous semantic routing graph and mark the virtual direct edges as hidden parallel links.

6. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 1, characterized in that: The specific process of monitoring the access status of heterogeneous links, analyzing protocol encapsulation depth and channel throughput, and calculating the ratio of channel throughput to protocol encapsulation depth to generate a channel bottleneck vector is as follows: Monitor the level transition events of the physical layer of the communication interface to identify the access status of abnormal links; The initial handshake message of the heterogeneous link is intercepted, and the network layer and data link layer header tags of the initial handshake message are decapsulated layer by layer. The total length of the nested headers is counted as the protocol encapsulation depth. Read the hardware configuration register of the network interface card corresponding to the heterogeneous link, and extract the maximum transmission rate locked by the hardware configuration register as the channel throughput. The channel throughput is mapped to an available bandwidth scalar, the protocol encapsulation depth is mapped to a transmission overhead scalar, the ratio of the available bandwidth scalar to the transmission overhead scalar is calculated, and a channel bottleneck vector containing the ratio, the available bandwidth scalar and the transmission overhead scalar, and the constraint relationship between the two is constructed.

7. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 6, characterized in that: The specific process of mapping the channel bottleneck vector to the heterogeneous semantic routing graph and updating the node connection weight values ​​is as follows: Extract the available bandwidth scalar, transmission overhead scalar, and ratio from the channel bottleneck vector, and locate the network topology node coordinates corresponding to the heterogeneous link in the heterogeneous semantic routing graph; Create new directed edges between the coordinates of network topology nodes, and combine the constraint relationship between the transmission overhead scalar and the available bandwidth scalar into a ratio to transform it into a network congestion penalty factor. The network congestion penalty factor is regularly fused with the original node connection weights in the heterogeneous semantic routing graph, and the node connection weights are updated in the newly added directed edges and heterogeneous semantic routing graph.

8. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 1, characterized in that: When the main acquisition link is interrupted, the maximum number of payload bytes corresponding to the channel bottleneck vector is obtained, and the specific process of extracting status telemetry metadata by pruning fields of the communication data stream based on the maximum number of payload bytes is as follows: Configure a heartbeat keep-alive timer to determine that the main acquisition link is interrupted when the heartbeat keep-alive timer times out and no response frame is received from the main acquisition link. The constraints contained in the channel bottleneck vector are analyzed, and the maximum transmission unit threshold of the physical channel supported by the constraints is retrieved as the maximum payload bytes. The message structure of the communication data stream is reverse-parsed based on the maximum payload byte count. As a field pruning step, regular system logs and descriptive text in the message structure are removed, and status telemetry metadata containing device operating parameters and fault codes in the message structure is extracted.

9. The method for topology modeling and heterogeneous link self-recovery of multi-protocol fusion acquisition terminals for data centers according to claim 8, characterized in that: The specific process of downgrading and reconstructing state telemetry metadata into low-bit-width protocol frames by combining a cross-protocol compatibility matrix, and sending low-bit-width protocol frames along the path with the minimum node connection weight value in heterogeneous semantic routing graphs and hidden parallel links is as follows: Query the cross-protocol compatibility matrix, match the source protocol identifier of the state telemetry metadata with the target protocol identifier in the cross-protocol compatibility matrix, and filter out the target low-width protocol template with the minimum semantic conversion overhead; The state telemetry metadata is encapsulated into the payload range of the target low-bit-width protocol template, and the communication checksum is recalculated to complete the downgrade reconstruction and generate a low-bit-width protocol frame. Within the global topology space of heterogeneous semantic routing graphs and hidden parallel links, the node connection weight values ​​are used as the pathfinding heuristic cost to perform shortest path retrieval and generate a recovery route sequence. Low-width protocol frames are then forwarded along the recovery route sequence.

Images