A method, device, equipment and medium for inspecting service parameters of an intelligent computing cluster

Abstract

The application discloses a kind of service parameter inspection methods, devices and equipment of intelligent computing cluster and medium, comprising: obtaining the service delivery information corresponding to intelligent computing cluster, according to service delivery information constructs business intent topology graph, and according to business intent topology graph collection multi-source configuration parameter;According to the complete execution link corresponding to intelligent computing cluster, the consistency check is carried out to the multi-source configuration parameter collected, and difference matrix is generated according to the check result;After service delivery acceptance, the business parameter associated with intelligent computing cluster is actively probed, and according to the active detection result and difference matrix, the abnormal reason of abnormal parameter is analyzed.The technical scheme of the embodiment of the application can identify the parameter inconsistency problem hidden in ROCE lossless network from the perspective of end-to-end path, thereby ensuring the consistency of business parameters of intelligent computing cluster from delivery acceptance to routine operation, improving the quality of intelligent computing service delivery, shortening the acceptance period and reducing the operation fault recovery time.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, equipment and medium for inspecting business parameters of an intelligent computing cluster. Background Technology

[0002] Currently, intelligent computing clusters commonly employ a lossless solution based on Remote Direct Memory Access over Converged Ethernet (ROCE) when supporting large-scale model training, inference, and storage acceleration services. This type of solution has strong consistency requirements for delivered service parameters. During project delivery, the network, computing, storage, and platform teams each deploy and configure their respective equipment, which can easily lead to end-to-end configuration drift due to differences in vendors, software versions, automated script branches, and manual changes. In subsequent operation and maintenance phases, equipment replacement, patch upgrades, service expansion, and partial cutovers can also introduce new inconsistencies.

[0003] When existing intelligent computing clusters inspect the delivered business parameters, they mostly only perform static checks on the configuration of a single device, lacking end-to-end consistency verification and failing to detect abnormal parameter combinations across devices. Secondly, they rely on manual spot checks or script-based comparisons of each device, resulting in long delivery cycles and low accuracy. Even if fault signs are detected, they often remain at the level of alarms or reports, failing to automatically generate a safe repair orchestration order based on changes in dependencies, resulting in poor flexibility in inspection methods. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and medium for inspecting the service parameters of an intelligent computing cluster. It can identify hidden parameter inconsistencies in the ROCE lossless network from an end-to-end path perspective, thereby ensuring the consistency of service parameters of the intelligent computing cluster from delivery and acceptance to daily operation and maintenance, improving the quality of intelligent computing service delivery, shortening the acceptance cycle, and reducing the operation and maintenance fault recovery time.

[0005] According to one aspect of the present invention, a method for inspecting service parameters of an intelligent computing cluster is provided, the method comprising: Obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology map based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology map; According to the complete execution link corresponding to the intelligent computing cluster, the consistency of the collected multi-source configuration parameters is verified, and a difference matrix is ​​generated based on the verification results. After the business is delivered and accepted, the business parameters associated with the intelligent computing cluster are actively probed, and the abnormal causes of the abnormal parameters are analyzed based on the active probe results and the difference matrix.

[0006] Optionally, the service delivery information includes multiple device nodes and attribute parameters corresponding to each device node; Construct a business intent topology map based on the business delivery information, including: A node set is constructed based on multiple device nodes in the service delivery information, and a communication path between each device node is generated based on the attribute parameters corresponding to each device node. The communication path is used as the connection edge between each device node. Based on the node set and the connection edge, a business intent topology graph is constructed.

[0007] Optionally, after collecting multi-source configuration parameters based on the business intent topology map, the method further includes: The multi-source configuration parameters are mapped to a unified field set, and a configuration parameter snapshot is generated based on the unified field set; The configuration parameter snapshot includes at least: timestamp, device version information, and parameter source identifier.

[0008] Optionally, according to the complete execution chain corresponding to the intelligent computing cluster, the collected multi-source configuration parameters are subjected to consistency verification, and a difference matrix is ​​generated based on the verification results, including: According to the complete execution link corresponding to the intelligent computing cluster, the actual configuration parameters collected in the configuration parameter snapshot are compared with the preset standard configuration parameters item by item; A difference matrix is ​​generated based on the comparison results; the difference matrix includes at least the field names, standard parameter values, actual parameter values, parameter influence range, risk level, and parameter dependencies of the differences.

[0009] Optionally, the service parameters associated with the intelligent computing cluster are actively probed, including: Initiate multiple test streams among the selected target devices and synchronously collect response data from each target device in response to the test streams; Based on the response data of each target device to the test stream, the active detection results corresponding to the intelligent computing cluster are generated.

[0010] Optionally, after analyzing the causes of anomalies in the abnormal parameters based on the active detection results and the difference matrix, the method further includes: Based on the active detection results and the difference matrix, analyze the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameters; According to a preset weight ratio, the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameter are weighted and summed to obtain the risk score corresponding to the abnormal parameter.

[0011] Optionally, after analyzing the causes of anomalies in the abnormal parameters based on the active detection results and the difference matrix, the method further includes: Based on the risk score corresponding to the abnormal parameter, the abnormal parameter is repaired, and after the repair is completed, the abnormal parameter is subjected to consistency verification and active detection again. Based on the new verification results and active detection results corresponding to the abnormal parameters, a verification report corresponding to the intelligent computing cluster is generated.

[0012] According to another aspect of the present invention, a service parameter inspection device for an intelligent computing cluster is provided, the device comprising: The topology graph construction module is used to obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology graph based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology graph. The difference verification module is used to perform consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, and generate a difference matrix based on the verification results. The active detection module is used to actively detect the business parameters associated with the intelligent computing cluster after the business is delivered and accepted, and to analyze the abnormal causes of abnormal parameters based on the active detection results and the difference matrix.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the business parameter inspection method of the intelligent computing cluster according to any embodiment of the present invention.

[0014] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being used to cause a processor to execute and implement the service parameter inspection method of the intelligent computing cluster according to any embodiment of the present invention.

[0015] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the business parameter inspection method for intelligent computing clusters as described in any embodiment of the present invention.

[0016] The technical solution provided by this invention obtains service delivery information corresponding to the intelligent computing cluster, constructs a service intent topology map based on the service delivery information, collects multi-source configuration parameters based on the service intent topology map, performs consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, generates a difference matrix based on the verification results, and actively probes the service parameters associated with the intelligent computing cluster after service delivery acceptance, and analyzes the causes of abnormal parameters based on the active probe results and the difference matrix. This technical means can identify parameter inconsistency problems hidden in the ROCE lossless network from an end-to-end path perspective, thereby ensuring the consistency of service parameters of the intelligent computing cluster from delivery acceptance to daily operation and maintenance, improving the quality of intelligent computing service delivery, shortening the acceptance cycle, and reducing the operation and maintenance fault recovery time.

[0017] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a business parameter inspection method for an intelligent computing cluster provided by an embodiment of the present invention; Figure 2 This is a flowchart of another method for inspecting service parameters of a smart computing cluster according to an embodiment of the present invention; Figure 3 This is a schematic diagram of intentional topology modeling provided according to an embodiment of the present invention; Figure 4 This is a flowchart of parameter acquisition and normalization processing provided by an embodiment of the present invention; Figure 5 This is a schematic diagram of a process for performing consistency verification and determining risk scores for business parameters in an intelligent computing cluster, according to an embodiment of the present invention. Figure 6 This is a flowchart of a closed-loop repair and rollback process for abnormal parameters provided by an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a business parameter inspection device for an intelligent computing cluster according to an embodiment of the present invention; Figure 8This is a schematic diagram of the structure of an electronic device that implements the business parameter inspection method of the intelligent computing cluster according to an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Figure 1 This is a flowchart illustrating a service parameter inspection method for an intelligent computing cluster, provided in an embodiment of the present invention. This embodiment is applicable to the inspection of service parameters delivered in an intelligent computing cluster. The method can be executed by a service parameter inspection device for the intelligent computing cluster, which can be implemented in hardware and / or software and configured in an electronic device, such as... Figure 1 As shown, the method includes: Step 110: Obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology map based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology map.

[0023] In this embodiment, optionally, a business intent topology map can be constructed based on the business parameters corresponding to each device in the business delivery information and the relationships between the devices. The business intent topology map defines the relationships between each business parameter and specific clusters, racks, container PODs, node roles, and business types.

[0024] Specifically, the multi-source configuration parameters include at least: the link bandwidth of the intelligent computing cluster, the maximum transmission unit (MTU) of the port, the priority-to-queue mapping relationship, the priority flow control (PFC) parameters, the minimum threshold of explicit congestion notification (ECN), the maximum threshold of ECN, the cache pool and shared buffer template, the parameters related to congestion notification priority (CNP), the differential services code point (DSCP) marking rules, the priority code point (PCP) marking rules, the load balancing strategy, the host network card driver version, the network card firmware version, the ROCE mode, the gateway identifier (GID) index, the communication test baseline indicators of the NVIDIA Collective Communications Library (NCCL) or the Message Passing Interface (MPI), and the change window constraint information.

[0025] Step 120: According to the complete execution link corresponding to the intelligent computing cluster, perform consistency verification on the collected multi-source configuration parameters, and generate a difference matrix based on the verification results.

[0026] In this embodiment, optionally, the actual collected multi-source configuration parameter Profile_actual can be compared with the preset standard configuration parameter Profile_target item by item according to the full path granularity of "host-access-aggregation-peer host" to obtain the difference matrix M.

[0027] Specifically, the discrepancy matrix includes at least the field names, target values, actual values, impact ranges, risk levels, and dependencies of the inconsistent fields. For parameters requiring end-to-end consistency, such as PFC priority, MTU, queue mapping parameters, the mapping relationship between DSCP and Traffic Control (TC), and ECN thresholds, this embodiment can perform strong consistency checks. For parameters with permissible fluctuations, such as firmware minor versions, cache availability, or link temperature, this embodiment can perform interval consistency checks.

[0028] After verifying the consistency of the collected multi-source configuration parameters using the above method, the path consistency score S_path can be calculated based on the verification results. S_path=Σ(w_i×c_i) / Σw_i, Where w_i represents the weight value corresponding to the i-th parameter item, and c_i represents the consistency check result corresponding to the parameter item (a value of 1 for complete consistency, 0.5 for partial consistency, and 0 for inconsistency). After obtaining the consistency score according to the above formula, the execution path status in the intelligent computing cluster can be divided into qualified, attention, risk, and blocked, to represent the degree of change of business parameters in the propagation path of the cluster.

[0029] Step 130: After the business is delivered and accepted, actively probe the business parameters associated with the intelligent computing cluster, and analyze the abnormal causes of the abnormal parameters based on the active probe results and the difference matrix.

[0030] In this embodiment, to avoid false alarms caused by the multi-source configuration parameters collected during the delivery phase alone, an active detection method is further proposed to be triggered in the window after delivery acceptance or change. That is, the target business parameters on the target devices in the intelligent computing cluster can be inspected regularly to obtain active detection results. Finally, based on the active detection results and the difference matrix, abnormal parameters in the intelligent computing cluster are identified and root cause analysis is performed on the abnormal parameters.

[0031] The technical solution provided by this invention obtains the service delivery information corresponding to the intelligent computing cluster, constructs a service intent topology map based on the service delivery information, collects multi-source configuration parameters based on the service intent topology map, performs consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, generates a difference matrix based on the verification results, and actively probes the service parameters associated with the intelligent computing cluster after service delivery acceptance, and analyzes the causes of abnormal parameters based on the active probe results and the difference matrix. This technical approach unifies the traditionally scattered four types of actions—"configuration inspection, performance acceptance, anomaly troubleshooting, and manual repair"—into a service intent-driven framework. It can identify parameter inconsistencies hidden in the ROCE lossless network from an end-to-end path perspective, thereby ensuring the consistency of service parameters of the intelligent computing cluster from delivery acceptance to daily operation and maintenance, improving the quality of intelligent computing service delivery, shortening the acceptance cycle, and reducing the operation and maintenance fault recovery time.

[0032] Figure 2 This is a flowchart illustrating another method for inspecting service parameters of an intelligent computing cluster provided by an embodiment of the present invention. In this embodiment, the service delivery information includes multiple device nodes and attribute parameters corresponding to each device node. Figure 2 As shown, the method includes: Step 210: Obtain the service delivery information corresponding to the intelligent computing cluster, construct a node set based on the multiple device nodes in the service delivery information, and generate the communication path between each device node based on the attribute parameters corresponding to each device node.

[0033] Step 220: Using the communication path as the connection edge between each device node, construct a service intent topology graph based on the node set and the connection edge, and collect multi-source configuration parameters based on the service intent topology graph.

[0034] In this embodiment, optionally, Figure 3 It can be a schematic diagram of an intentional topology model, such as Figure 3 As shown, based on the path relationships formed by the training server, inference server, storage node, TOR switch, Leaf switch, and Spine switch, as well as the target fields involved in the delivery intent template, such as MTU, PFC priority, ECN threshold, queue mapping, DSCP / PCP tag, and network card parameters, different service parameters can be bound to different paths or node roles.

[0035] In a specific embodiment, after receiving service delivery information, a service intent topology graph G=(V,E) can be constructed based on the project delivery list, equipment ledger, networking method, service type, target performance indicators, and maintenance constraints in the service delivery information. Here, the node set V includes training servers, inference servers, storage nodes, TOR switches, Leaf switches, Spine switches, and management nodes; the edge set E represents the actual communication links or logical forwarding paths. This embodiment can hierarchically label the graph structure according to training domain, tenant, service plane, and fault domain, and generate a target parameter profile (Profile_target) for each path.

[0036] Step 230: Map the multi-source configuration parameters to a unified field set, and generate a configuration parameter snapshot based on the unified field set; wherein the configuration parameter snapshot includes at least: timestamp, device version information, and parameter source identifier.

[0037] In this embodiment, Figure 4 This can be a flowchart of parameter acquisition and normalization processing, such as... Figure 4 As shown, raw configuration data can be obtained from switch interfaces, host agents, automation platforms, and configuration snapshot libraries, respectively. Fields from different vendors and versions are parsed, standardized, mapped, timestamped, and marked with trustworthiness, ultimately forming a unified configuration profile, Profile_actual.

[0038] In one specific embodiment, the configuration parameters collected from the switch include at least the Quality of Service (QoS) template, PFC / ECN configuration parameters, queue scheduling parameters, buffer pool threshold, link aggregation configuration, Virtual Local Area Network (VLAN) / Virtual Extensible LAN (VXLAN) mapping, and port status; the configuration parameters collected from the host include at least the network card driver version, firmware version, ROCE switch status, PFC Trust mode, jumbo frame parameters, CPU affinity policy, GID entries, and RDMA communication library version.

[0039] Step 240: According to the complete execution link corresponding to the intelligent computing cluster, compare the actual configuration parameters collected in the configuration parameter snapshot with the preset standard configuration parameters item by item, and generate a difference matrix based on the comparison results.

[0040] In this embodiment, the difference matrix includes at least the field names with differences, standard parameter values, actual parameter values, parameter influence range, risk level, and parameter dependencies.

[0041] Step 250: After the service delivery and acceptance, initiate multiple test flows among the selected target devices, and synchronously collect the response data of each target device to the test flow. Based on the response data of each target device to the test flow, generate the active probing results corresponding to the intelligent computing cluster.

[0042] In this embodiment, to avoid false alarms caused by relying solely on configuration snapshots, a proactive probing method is further proposed to be triggered within the delivery acceptance or post-change window. Optionally, multiple types of test flows can be initiated between selected host pairs, including at least small message high-frequency control flows, large packet burst flows, mixed read / write flows, and many-to-many concurrent flows, while simultaneously collecting real-time count values ​​from the switch and hosts. The count values ​​include at least the number of PFC pause frames, the number of ECN tags, the number of CNP packets, queue dwell length, port packet loss, retransmission count, link utilization, latency jitter, and NCCL / MPI throughput metrics.

[0043] The advantage of this setup is that proactive detection allows for rapid and effective acceptance of delivered business parameters in intelligent computing cluster delivery scenarios; and rapid detection of business parameters in intelligent computing cluster operation and maintenance scenarios, reducing fault recovery time during the operation and maintenance phase and improving operation and maintenance efficiency.

[0044] Step 260: Analyze the causes of abnormal parameters based on the active detection results and the difference matrix.

[0045] Step 270: Based on the active detection results and the difference matrix, analyze the drift degree, anomaly intensity, impact range and disturbance cost of the repair action corresponding to the abnormal parameter. According to the preset weight ratio, perform a weighted summation of the drift degree, anomaly intensity, impact range and disturbance cost of the repair action corresponding to the abnormal parameter to obtain the risk score corresponding to the abnormal parameter.

[0046] In this embodiment, optionally, Figure 5 This can be a flowchart illustrating the process of verifying the consistency of business parameters in an intelligent computing cluster and determining risk scores. It can compare the actual configuration profile (Profile_actual) with the target configuration profile (Profile_target) item by item according to the end-to-end path, generating a difference matrix M. Then, it combines the results of active detection and real-time telemetry indicators to calculate the path consistency score S_path and the risk score R, subsequently outputting classification results such as qualified, concerning, risky, or blocked.

[0047] In a specific embodiment, a unified time window can be used to perform correlation calculations on configuration snapshots, active probing results, and real-time telemetry. For example, when "PFC pause frames continue to increase and ECN tags are insufficient" is detected, the anomaly is determined to be a high probability problem of missing ECN thresholds, excessively high thresholds, or incorrect priority mapping; when "CNP numbers increase abnormally and throughput decreases" is detected, the anomaly is determined to be due to congestion feedback hypersensitivity, buffer configuration mismatch, or unreasonable queue weights; when "end-to-end MTU inconsistency and message timeouts or fragmentation occur" is detected, the anomaly is determined to be due to host and switch jumbo frame configuration drift; when "a specific link is continuously heavily occupied while other equivalent links are idle" is detected, the anomaly is determined to be due to abnormal load balancing hash strategy or path reachability configuration.

[0048] Optionally, the formula for determining the risk score for the identified abnormal parameters is as follows: R = αD + βT + γB + δA Where D represents the degree of drift corresponding to the abnormal parameter, T represents the abnormality intensity, B represents the range of influence, A represents the disturbance cost of the repair action, and α, β, γ, and δ represent the preset weight values, respectively.

[0049] Step 280: Repair the abnormal parameters according to the risk score corresponding to the abnormal parameters, and perform consistency verification and active detection on the abnormal parameters again after the repair is completed. Generate a backtesting report corresponding to the intelligent computing cluster based on the new verification results and active detection results corresponding to the abnormal parameters.

[0050] In this embodiment, optionally, high-risk abnormal parameters can be prioritized for repair based on the dependency topology relationship between the risk score and the parameters. Figure 6This is a flowchart of the closed-loop repair and rollback process for abnormal parameters in this embodiment. A repair dependency directed graph can be generated according to the difference type and dependency topology, and the pre-check, single link grayscale, single POD ramp-up and global convergence stages can be executed in sequence. In each stage, the process is determined based on indicators such as throughput, latency, PFC pause frame, ECN mark and job status to determine whether to enter the next stage or trigger automatic rollback.

[0051] In a specific implementation, a directed graph of repair dependencies can be constructed after identifying abnormal parameters. For example, for changes involving both switches and hosts, a multi-stage repair sequence of "pre-check - single-link canary deployment - single POD ramp-up - global convergence" is generated first. Each stage defines preconditions, concurrency limits, protection thresholds, and automatic rollback conditions. For instance, when repairing a PFC priority mapping error, canary configuration can be performed first on backup or low-load links. Only after actively probing and confirming a decrease in PFC pause frames, recovery of throughput, and no new job timeouts are added, can the rollback be extended to similar devices. If, after extension, key metrics exceed thresholds, a rollback to the previous version configuration snapshot is performed in reverse order: first hosts then switches, or first edge then core.

[0052] After the repair is completed using the above methods, consistency checks and proactive probing can be automatically performed again to generate a delivery acceptance report or an operation and maintenance retest report. The retest report should include at least the differences before and after the repair, a comparison of key indicators, risk mitigation results, scope of impact, and a traceable configuration version number. For verified abnormal parameters, the "abnormal characteristics - root cause type - repair action - retest result" structure can be processed into knowledge entries for direct reuse in subsequent similar projects, thereby improving the standardization of delivery and the automation of operation and maintenance.

[0053] In one embodiment of this example, a security constraint mechanism can be set for the intelligent computing cluster: when it is detected that the intelligent computing cluster is in a high-load training window, the link redundancy is insufficient, the target device has a hardware alarm, or the configuration change does not meet the approval conditions, only risk suggestions and candidate repair scripts are output, and the issuance action is not automatically executed, so as to avoid introducing greater disturbances during peak production periods.

[0054] The technical solution provided in this invention involves obtaining service delivery information corresponding to an intelligent computing cluster, constructing a node set based on multiple device nodes in the service delivery information, generating communication paths between device nodes based on the attribute parameters corresponding to each device node, using the communication paths as connection edges between device nodes, constructing a service intent topology graph based on the node set and connection edges, collecting multi-source configuration parameters based on the service intent topology graph, mapping the multi-source configuration parameters to a unified field set, generating a configuration parameter snapshot based on the unified field set, comparing the actual configuration parameters collected in the configuration parameter snapshot with preset standard configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, generating a difference matrix based on the comparison results, and initiating multiple types of test flows between selected target devices after service delivery acceptance, while simultaneously collecting various... The system utilizes the response data of target devices to the test flow to generate active detection results for the intelligent computing cluster. Based on these responses, it analyzes the causes of anomalies in abnormal parameters using the active detection results and a difference matrix. Then, it weights and sums the drift degree, anomaly intensity, impact range, and disturbance cost of repair actions for each anomaly parameter according to preset weight ratios, resulting in a risk score. The system then repairs the anomaly parameters and performs consistency verification and active detection again after repair. Finally, it generates a feedback report for the intelligent computing cluster based on the new verification and active detection results. This technical approach forms a closed-loop mechanism from consistency identification, risk scoring, root cause classification to tiered repair and effect feedback, improving the quality of intelligent computing service delivery, shortening the acceptance cycle, and reducing maintenance and fault recovery time.

[0055] Figure 7 This is a schematic diagram of the structure of a service parameter inspection device for an intelligent computing cluster provided in an embodiment of the present invention, as shown below. Figure 7 As shown, the device includes: a topology graph construction module 710, a difference verification module 720, and an active detection module 730.

[0056] The topology graph construction module 710 is used to obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology graph based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology graph. The difference verification module 720 is used to perform consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, and generate a difference matrix based on the verification result. The active detection module 730 is used to actively detect the business parameters associated with the intelligent computing cluster after the business is delivered and accepted, and to analyze the abnormal causes of abnormal parameters based on the active detection results and the difference matrix.

[0057] The technical solution provided by this invention obtains service delivery information corresponding to the intelligent computing cluster, constructs a service intent topology map based on the service delivery information, collects multi-source configuration parameters based on the service intent topology map, performs consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, generates a difference matrix based on the verification results, and actively probes the service parameters associated with the intelligent computing cluster after service delivery acceptance, and analyzes the causes of abnormal parameters based on the active probe results and the difference matrix. This technical means can identify parameter inconsistency problems hidden in the ROCE lossless network from an end-to-end path perspective, thereby ensuring the consistency of service parameters of the intelligent computing cluster from delivery acceptance to daily operation and maintenance, improving the quality of intelligent computing service delivery, shortening the acceptance cycle, and reducing the operation and maintenance fault recovery time.

[0058] Based on the above embodiments, the service delivery information includes multiple device nodes and attribute parameters corresponding to each device node.

[0059] Topology graph construction module 710 includes: The intent topology construction unit is used to construct a node set based on multiple device nodes in the service delivery information, generate communication paths between device nodes based on the attribute parameters corresponding to each device node, use the communication paths as connection edges between device nodes, and construct a service intent topology graph based on the node set and connection edges. The parameter mapping unit is used to map the multi-source configuration parameters to a unified field set and generate a configuration parameter snapshot based on the unified field set; wherein the configuration parameter snapshot includes at least: timestamp, device version information and parameter source identifier.

[0060] The difference verification module 720 includes: The parameter comparison unit is used to compare the actual configuration parameters collected in the configuration parameter snapshot with the preset standard configuration parameters according to the complete execution link corresponding to the intelligent computing cluster; generate a difference matrix based on the comparison results; the difference matrix includes at least the field names with differences, standard parameter values, actual parameter values, parameter impact range, risk level, and parameter dependencies.

[0061] The active detection module 730 includes: The test flow triggering unit is used to initiate multiple types of test flows among selected target devices and synchronously collect the response data of each target device to the test flow; based on the response data of each target device to the test flow, it generates the active probing results corresponding to the intelligent computing cluster. The risk scoring unit is used to analyze the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameter based on the active detection results and the difference matrix; and to perform a weighted summation of the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameter according to a preset weight ratio to obtain the risk score corresponding to the abnormal parameter. The parameter repair unit is used to repair the abnormal parameter according to the risk score corresponding to the abnormal parameter, and after the repair is completed, to perform consistency verification and active detection on the abnormal parameter again; and to generate a backtesting report corresponding to the intelligent computing cluster based on the new verification result and active detection result corresponding to the abnormal parameter.

[0062] The above-described apparatus can execute the methods provided in all the foregoing embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the above methods. Technical details not described in detail in the embodiments of the present invention can be found in the methods provided in all the foregoing embodiments of the present invention.

[0063] Figure 8 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0064] like Figure 8 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) or a random access memory (RAM), communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 12 or loaded from the storage unit 18 into the random access memory 13. The random access memory 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, the read-only memory 12, and the random access memory 13 are interconnected via a bus 14. Input / output (I / O) interfaces are also connected to the bus 14.

[0065] Multiple components in electronic device 10 are connected to input / output interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0066] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processing (DSP) processors, and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the business parameter inspection method of an intelligent computing cluster.

[0067] In some embodiments, the service parameter inspection method for the intelligent computing cluster can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via read-only memory 12 and / or communication unit 19. When the computer program is loaded into random access memory 13 and executed by processor 11, one or more steps of the service parameter inspection method for the intelligent computing cluster described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the service parameter inspection method for the intelligent computing cluster by any other suitable means (e.g., by means of firmware).

[0068] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chips (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0069] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0070] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0071] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD)) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0072] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0073] A computing system can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system. It addresses the shortcomings of traditional physical hosts and Virtual Private Servers (VPS) in terms of management difficulty and weak business scalability.

[0074] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0075] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for inspecting business parameters of an intelligent computing cluster, characterized in that, The method includes: Obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology map based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology map; According to the complete execution link corresponding to the intelligent computing cluster, the consistency of the collected multi-source configuration parameters is verified, and a difference matrix is ​​generated based on the verification results. After the business is delivered and accepted, the business parameters associated with the intelligent computing cluster are actively probed, and the abnormal causes of the abnormal parameters are analyzed based on the active probe results and the difference matrix.

2. The method according to claim 1, characterized in that, The service delivery information includes multiple device nodes and attribute parameters corresponding to each device node; Construct a business intent topology map based on the business delivery information, including: A node set is constructed based on multiple device nodes in the service delivery information, and a communication path between each device node is generated based on the attribute parameters corresponding to each device node. The communication path is used as the connection edge between each device node. Based on the node set and the connection edge, a business intent topology graph is constructed.

3. The method according to claim 1, characterized in that, After collecting multi-source configuration parameters based on the business intent topology map, the process also includes: The multi-source configuration parameters are mapped to a unified field set, and a configuration parameter snapshot is generated based on the unified field set; The configuration parameter snapshot includes at least: timestamp, device version information, and parameter source identifier.

4. The method according to claim 3, characterized in that, According to the complete execution chain corresponding to the intelligent computing cluster, the collected multi-source configuration parameters are subjected to consistency verification, and a difference matrix is ​​generated based on the verification results, including: According to the complete execution link corresponding to the intelligent computing cluster, the actual configuration parameters collected in the configuration parameter snapshot are compared with the preset standard configuration parameters item by item; A difference matrix is ​​generated based on the comparison results; the difference matrix includes at least the field names, standard parameter values, actual parameter values, parameter influence range, risk level, and parameter dependencies of the differences.

5. The method according to claim 1, characterized in that, Actively probe the service parameters associated with the intelligent computing cluster, including: Initiate multiple test streams among the selected target devices and synchronously collect response data from each target device in response to the test streams; Based on the response data of each target device to the test stream, the active detection results corresponding to the intelligent computing cluster are generated.

6. The method according to claim 1, characterized in that, After analyzing the causes of anomalies in the abnormal parameters based on the active detection results and the difference matrix, the process also includes: Based on the active detection results and the difference matrix, analyze the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameters; According to a preset weight ratio, the drift degree, anomaly intensity, impact range, and disturbance cost of the repair action corresponding to the abnormal parameter are weighted and summed to obtain the risk score corresponding to the abnormal parameter.

7. The method according to claim 6, characterized in that, After analyzing the causes of anomalies in the abnormal parameters based on the active detection results and the difference matrix, the process also includes: Based on the risk score corresponding to the abnormal parameter, the abnormal parameter is repaired, and after the repair is completed, the abnormal parameter is subjected to consistency verification and active detection again. Based on the new verification results and active detection results corresponding to the abnormal parameters, a verification report corresponding to the intelligent computing cluster is generated.

8. A service parameter inspection device for an intelligent computing cluster, characterized in that, The device includes: The topology graph construction module is used to obtain the business delivery information corresponding to the intelligent computing cluster, construct a business intent topology graph based on the business delivery information, and collect multi-source configuration parameters based on the business intent topology graph. The difference verification module is used to perform consistency verification on the collected multi-source configuration parameters according to the complete execution link corresponding to the intelligent computing cluster, and generate a difference matrix based on the verification results. The active detection module is used to actively detect the business parameters associated with the intelligent computing cluster after the business is delivered and accepted, and to analyze the abnormal causes of abnormal parameters based on the active detection results and the difference matrix.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the business parameter inspection method of the intelligent computing cluster as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which are used to cause the processor to execute the business parameter inspection method of the intelligent computing cluster as described in any one of claims 1-7.

Images