Computing power network state monitoring method and device, computer device and storage medium

Abstract

This invention belongs to the field of quantum communication and relates to a method, device, computer equipment, and storage medium for monitoring the state of computing networks. The method includes: defining and configuring a situational awareness system, mapping quantum hardware states and network behavior to quantitative indicators with clear physical meaning and computational methods; collecting raw data required for calculating situational awareness from various quantum nodes, network links, and control systems based on the situational awareness system; performing real-time calculation and multi-level fusion of situational awareness from the raw data; analyzing the intrinsic correlation between situational awareness and the laws governing their evolution over time to obtain analysis results; performing dynamic risk assessment and graded early warning of the network based on the analysis results; and performing adaptive scheduling and closed-loop optimization of the network state based on the results of the dynamic risk assessment and graded early warning. A multi-dimensional situational awareness system is constructed, overcoming the limitations of single-dimensional assessment; real-time calculation and deep fusion of situational awareness are achieved, improving the dynamism and accuracy of monitoring.

Description

Technical Field

[0001] This invention relates to the field of quantum communication technology, and in particular to methods, devices, computer equipment, and storage media for monitoring the status of computing networks. Background Technology

[0002] With the rapid development of quantum computing technology, distributed quantum computing (DQC) has become an important way to overcome the physical limitations of a single quantum device.

[0003] In the DQC architecture, multiple quantum processing units (QPUs) are interconnected through a quantum network to jointly complete complex quantum computing tasks.

[0004] However, unlike traditional quantum computing, situational awareness in DQC requires the simultaneous tracking of dynamic indicators across four dimensions: the computing power status of multiple nodes, the performance of the quantum network, the effectiveness of distributed error correction, and the progress of task execution. These indicators not only need to reflect the current state but also predict future trends, reveal system correlations, and assess potential risks.

[0005] Currently, several simulation testing platforms have emerged in the field of quantum computing, such as QTest, iQuantum, and NetSquid. They mainly focus on single-dimensional performance evaluation and lack the ability to comprehensively analyze multi-dimensional state variables.

[0006] In addition, the existing system has significant shortcomings in real-time monitoring, dynamic scheduling and risk prediction, and cannot meet the actual needs of large-scale DQC networks. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a method for monitoring the status of a computing network, employing the following technical solution, including the following steps:

[0008] Define and configure the state quantity system to map quantum hardware states and network behaviors into quantitative indicators with clear physical meaning and computational methods;

[0009] Based on the aforementioned state quantity system, raw data required for calculating the state quantity are collected from various quantum nodes, network links, and control systems;

[0010] The raw data is subjected to real-time calculation of situational parameters and multi-level fusion.

[0011] Analyze the intrinsic relationships between state variables and the patterns of their evolution over time to obtain analytical results;

[0012] Based on the analysis results, network dynamic risk assessment and graded early warning are conducted.

[0013] Based on dynamic risk assessment and hierarchical early warning results, adaptive scheduling and closed-loop optimization of network status are carried out.

[0014] Preferably, the step of defining and configuring the state quantity system, which maps quantum hardware states and network behavior into quantitative indicators with clear physical meaning and computational methods, specifically includes:

[0015] Standardize the definitions of core state variables in the dimensions of node computing power, quantum network, distributed error correction, and task execution.

[0016] Configure differentiated correction formulas for different hardware routes for core status variables;

[0017] Multiple levels of early warning thresholds are preset for core situational variables.

[0018] Preferably, the step of collecting the raw data required for calculating the state quantity from various quantum nodes, network links, and control systems based on the state quantity system specifically includes:

[0019] Based on the aforementioned situational quantity system, a distributed acquisition terminal is deployed to achieve standardized encapsulation of hardware interfaces and collect the raw data required for calculating situational quantities.

[0020] Based on gRPC, the collected raw data is synchronized with low latency.

[0021] The original data is stored based on erasure coding.

[0022] Preferably, the step of performing real-time situational quantity calculation and multi-level fusion on the original data specifically includes:

[0023] Based on a high-performance computing chip, the raw data is subjected to real-time parallel computation.

[0024] Real-time feedback of continuous quantum error correction based on FPGA;

[0025] By using data fusion, we can achieve three-layer information fusion: feature-understanding-evaluation.

[0026] Preferably, the steps for obtaining analysis results by analyzing the intrinsic relationships between the analytical state variables and their evolution over time specifically include:

[0027] Based on Bayesian networks, we perform situational correlation modeling.

[0028] Time series prediction based on hybrid quantum classical LSTM;

[0029] System behavior modeling is performed based on Markov decision processes.

[0030] Preferably, the step of conducting network dynamic risk assessment and graded early warning based on the analysis results specifically includes:

[0031] Based on the analysis results, a multivariate risk scoring model is constructed;

[0032] Map continuous risk scores to warning levels and define response strategies for different levels;

[0033] Generate a standardized risk assessment report.

[0034] Preferably, the step of performing network state adaptive scheduling and closed-loop optimization based on dynamic risk assessment and hierarchical early warning results specifically includes:

[0035] Based on dynamic risk assessment and hierarchical early warning results, real-time scheduling optimization is performed through deep reinforcement learning;

[0036] By linking tiered early warning with intelligent response, an automatic closed-loop response to early warning signals can be achieved.

[0037] By evaluating the effectiveness of implemented decisions and using newly generated data for continuous training and optimization.

[0038] To address the aforementioned technical problems, the present invention also provides a computing power network status monitoring device, which employs the following technical solution, including:

[0039] The definition module is used to define and configure the state quantity system, mapping the quantum hardware state and network behavior into quantitative indicators with clear physical meaning and computational methods;

[0040] The acquisition module is used to acquire the raw data required for calculating the state variables from various quantum nodes, network links, and control systems based on the state variable system.

[0041] The fusion module is used to perform real-time calculation of situational parameters and multi-level fusion of the raw data;

[0042] The analysis module is used to analyze the intrinsic relationships between state variables and their evolution over time to obtain analysis results.

[0043] The assessment module is used to perform dynamic network risk assessment and graded early warning based on the analysis results.

[0044] The optimization module is used to perform adaptive scheduling and closed-loop optimization of network status based on dynamic risk assessment and hierarchical early warning results.

[0045] To address the aforementioned technical problems, the present invention also provides a computer device that employs the technical solution described below, comprising a memory and a processor. The memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the aforementioned computing power network status monitoring method.

[0046] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium, which employs the technical solution described below. The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the aforementioned computing network status monitoring method.

[0047] Compared with the prior art, the present invention has the following main advantages:

[0048] (1) By defining and configuring the situation quantity system, the four dimensions of quantum hardware state, network performance, error correction effect and task progress are uniformly quantitatively modeled, realizing the all-round situational awareness of the DQC system from the physical layer to the task layer, providing a complete data foundation for the understanding and control of complex systems, constructing a multi-dimensional situational awareness system, and breaking through the limitations of single-dimensional evaluation.

[0049] (2) By performing multi-level fusion of multi-source heterogeneous data, it is possible not only to dynamically track changes in system state, but also to explore the intrinsic correlation and evolution law between situational quantities, thereby shifting from passive monitoring to active cognition, providing high-quality real-time intelligence for subsequent decision-making, realizing real-time calculation and deep fusion of situational quantities, and improving the dynamics and accuracy of monitoring.

[0050] (3) Based on situational analysis, a dynamic risk assessment and hierarchical early warning mechanism is introduced, which can identify systemic risks and issue early warnings. More importantly, through the closed-loop process of perception-analysis-assessment-scheduling, adaptive scheduling and continuous optimization of network status are realized, which significantly improves the robustness and execution efficiency of DQC network in dynamic environment, and has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control. Attached Figure Description

[0051] To more clearly illustrate the solutions in this invention, the accompanying drawings used in the description of the embodiments of this invention will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0052] Figure 1 This is a flowchart of an embodiment of the computing power network status monitoring method of the present invention;

[0053] Figure 2 This is a schematic diagram of the structure of an embodiment of the computing power network status monitoring device of the present invention;

[0054] Figure 3 This is a schematic diagram of another embodiment of the computing power network status monitoring device of the present invention;

[0055] Figure 4 This is a schematic diagram of the structure of an embodiment of the computer device of the present invention. Detailed Implementation

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects and not to describe a particular order.

[0057] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0058] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0059] It should be noted that the computing power network status monitoring method provided in the embodiments of the present invention is generally executed by a server / terminal device, and correspondingly, the computing power network status monitoring device is generally installed in the server / terminal device.

[0060] It should be understood that the number of terminal devices, networks, and servers is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used.

[0061] Example 1

[0062] Please refer to Figure 1 A flowchart of an embodiment of the computing power network status monitoring method of the present invention is shown. The computing power network status monitoring method includes the following steps:

[0063] Step S1: Define and configure the state quantity system, mapping the quantum hardware state and network behavior into quantitative indicators with clear physical meaning and computational methods.

[0064] In this embodiment, the electronic device (e.g., a server / terminal device) on which the computing power network status monitoring method runs can receive computing power network status monitoring requests via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra wideband) connections, and other currently known or future-developed wireless connection methods.

[0065] In this embodiment, step S1 may specifically include the following steps:

[0066] S11 standardizes the definition of core state variables for the node computing power dimension, quantum network dimension, distributed error correction dimension, and task execution dimension.

[0067] Establish a multi-dimensional classification and coding system. Specifically, firstly, divide the DQC system state into four dimensions:

[0068] The node computing power dimension is used to characterize the computing power and stability of each quantum processing unit (QPU).

[0069] The quantum network dimension is used to describe the quality and efficiency of quantum communication links between QPUs.

[0070] The distributed error correction dimension is used to measure the performance and overhead of error correction mechanisms deployed to combat noise.

[0071] The task execution dimension is used to track the progress, efficiency, and reliability of upper-layer application tasks. Based on this, several core state variables are extracted (such as dynamic quantum volume, entanglement distribution success rate, logic error rate, and subtask completion progress), and each state variable is given a standardized definition.

[0072] For example:

[0073] Dynamic quantum volume ( : Defined as the complexity of the largest random quantum circuit that a quantum processor can reliably execute within a specific time window at a given noise level, it is a comprehensive indicator of the real-time availability of node computing power.

[0074] Entanglement distribution success rate ( : Defined as the ratio of the number of times two QPUs successfully establish entangled pairs to the total number of distribution attempts within a certain period of time, it is a core indicator reflecting the reliability of quantum links.

[0075] Logical error rate ( : Defined as the probability that a logical qubit will make an error when performing an operation after being encoded by quantum error correction (QEC), it is a key indicator for measuring the effectiveness of distributed error correction.

[0076] Subtask completion progress ( ): Defined as the first In a subtask, the ratio of the number of quantum gates that have been successfully executed to the total number of quantum gates required to be executed in that subtask is the most intuitive reflection of the task's execution status.

[0077] This standardized definition transforms the ambiguous system state into quantifiable, comparable, and traceable objective data.

[0078] The purpose of step S11 is to systematically divide the operating state of the DQC system into four interrelated but independent dimensions, starting from the top-level design: node computing power, quantum network, distributed error correction, and task execution. It also defines the core status variables for each dimension to ensure the comprehensiveness and systematic nature of the monitoring system.

[0079] S12 is configured with differentiated correction formulas for core status variables based on different hardware routes.

[0080] The general calculation formula is modified based on hardware characteristic parameters. This is done using dynamic quantum volume (…). Taking as an example, its general calculation formula can be expressed as:

[0081] .in, It is the number of currently available and effective qubits. This is the maximum allowable circuit depth under the current noise level. In specific implementations, and The calculation requires the introduction of hardware-related parameters for correction.

[0082] For superconducting quantum computing nodes, the coherence time of a qubit ( The coherence time decay factor is a key factor limiting its availability. Therefore, the calculation of the number of available qubits requires the introduction of this factor, and the corrected formula is:

[0083] .

[0084] in, It is the total number of physical qubits on the node. It is the first The current fidelity of each quantum bit, It is a preset fidelity threshold (which can be set to 0.9-0.99). It is an indicator function (1 if the condition is met, 0 otherwise). It is the time interval since the last calibration. It is the coherence time of that quantum bit.

[0085] This formula quantifies the impact of decoherence on the decrease in qubit availability over time.

[0086] For optical quantum computing nodes, photon loss during transmission and storage is a major issue. Therefore, when calculating the effective number of qubits, the photon survival rate (PRR) needs to be considered. Corrections will be made:

[0087] .

[0088] This formula shows that even if a photon qubit has a high fidelity, it cannot be considered a reliable and usable resource if its survival probability is low.

[0089] By pre-setting multiple sets of hardware correction formulas for each state variable and selecting and configuring parameters according to the actual deployed hardware type during the system initialization phase, this method achieves good adaptability to different quantum technology routes.

[0090] Step S12 aims to address the heterogeneity issue of quantum hardware approaches (superconducting, photonic, ion trap, topological, etc.). Since qubits implemented using different physical methods possess drastically different physical characteristics and noise models, directly using a general formula can lead to inaccurate state quantity calculations. Therefore, it is necessary to configure differentiated correction formulas for each core state quantity based on different hardware approaches to ensure the accuracy of monitoring data and the universality of the solution.

[0091] S13 sets multiple levels of early warning thresholds for core status variables.

[0092] Combining theoretical analysis, historical data, and expert experience, a set of tiered thresholds (Th) is set for each state variable. For example, for quantum gate operation fidelity (Th... Three levels of warning thresholds can be set:

[0093] Note the threshold ( ):For example When the fidelity falls below this value, the system records a notice event, prompting operations and maintenance personnel to pay attention to the performance fluctuations of that node, but will not trigger automatic intervention.

[0094] Warning threshold ( ):For example When the fidelity falls below this value, the system triggers an alert, notifies operations and maintenance personnel to intervene and check, and may allocate more fault-tolerant resources to new tasks on that node.

[0095] Danger threshold ( ):For example When the fidelity falls below this value, the system determines that the node is unreliable and will automatically trigger an emergency response, such as pausing the assignment of new tasks to the node or migrating low-priority tasks currently running on the node.

[0096] These thresholds are not static; they can be dynamically adjusted based on historical system data. For example, using a sliding window statistical method, dynamic thresholds can be calculated based on recent performance data to more sensitively capture abnormal patterns. The preset threshold system provides clear boundary conditions for the system's automated decision-making.

[0097] The purpose of step S13 is to provide quantitative evaluation criteria for subsequent status monitoring and risk assessment. By presetting multiple levels of early warning thresholds for each core status variable, the system can automatically determine whether the current status is normal, abnormal, or dangerous, thereby achieving a leap from data monitoring to status perception.

[0098] The purpose of step S1 is to address the fundamental problem of how to describe and quantify the state of distributed quantum computing (DQC) networks. Its function is to establish a standardized, structured, and scalable monitoring framework that maps the complex, heterogeneous underlying quantum hardware states and network behaviors into a series of quantitative indicators with clear physical meaning and computational methods, thereby providing a unified language and data foundation for all subsequent monitoring, analysis, and control activities.

[0099] Step S2: Based on the state quantity system, collect the raw data required for calculating the state quantity from various quantum nodes, network links and control systems.

[0100] In this embodiment, step S2 may specifically include the following steps:

[0101] S21. Based on the aforementioned situational quantity system, deploy distributed acquisition terminals to achieve standardized encapsulation of hardware interfaces and collect the raw data required for calculating situational quantities.

[0102] An edge-to-edge collaborative acquisition architecture is adopted, and standardized encapsulation of hardware interfaces is implemented.

[0103] An edge acquisition terminal is deployed near each QPU. This terminal can be dedicated hardware with an FPGA or a high-performance software agent. It connects directly to the QPU's cryogenic control layer or real-time controller to capture the lowest-level physical events. For example, for superconducting quantum nodes, the FPGA real-time controller connects directly to the qubit manipulation and readout circuitry via the AXI4-Stream protocol to acquire the raw analog-to-digital (ADC) data from quantum state readouts at high speed. For quantum network links, dedicated network probes are deployed to capture the state of devices such as entanglement swaps and quantum repeaters.

[0104] To ensure compatibility with hardware from different manufacturers, the acquisition terminal needs to implement multiple hardware interface protocols. For example, for superconducting devices conforming to the YD / T 6060-2024 standard, a data receiving interface is developed based on the SwiftNIO non-blocking I / O framework to achieve efficient data reading. For photonic quantum devices, the output of the photon counting detector is accurately read through frequency-converting waveguides and OAM (orbital angular momentum) encoded interfaces to measure microscopic parameters such as photon survival rate and entanglement generation rate. Regardless of the underlying interface, the acquisition terminal ultimately encapsulates the raw data into a unified data packet. This data packet contains core fields such as qubit ID, global timestamp, operation type (e.g., X-gate, CX-gate, measurement), and measurement result (e.g., fidelity value), laying the foundation for unified processing at the upper layer.

[0105] Step S21 serves to ensure physical access to the required data source. In a DQC environment, the data source is highly distributed and heterogeneous, including the controller within each QPU, the quantum and classical channels connecting the QPUs, and the management nodes running and scheduling tasks. Therefore, acquisition terminals need to be deployed at various key locations, and adapters need to be developed for different hardware interfaces.

[0106] S22, based on gRPC, perform low-latency data synchronization on the collected raw data.

[0107] It combines a hardware time synchronization protocol with a high-performance remote procedure call (RPC) framework.

[0108] Employing a Precision Time Protocol (PTP) network such as White Rabbit (WR), sub-nanosecond time synchronization accuracy is provided for the master clocks of all acquisition terminals and the data center. Each acquisition terminal is timestamped by WR hardware when generating data packets, ensuring that events across all nodes are based on the same timeline.

[0109] In the control plane, gRPC (an RPC framework based on HTTP / 2) is used as the primary communication protocol between each acquisition terminal and the central data processing center. gRPC utilizes Protocol Buffers as its interface definition language and data serialization protocol, possessing efficient binary serialization capabilities and multiplexing features, significantly reducing data transmission latency and network overhead. The acquisition terminals push data packets with precise timestamps to the data center in real time via gRPC's streaming function, achieving efficient data aggregation.

[0110] The purpose of step S22 is to resolve the time consistency problem of data among distributed nodes. Quantum operations (such as entanglement distribution) operate on the timescale of nanoseconds to microseconds. If the timestamps of data from different nodes are inconsistent, subsequent correlation analysis (such as calculating end-to-end communication latency) will be meaningless. Therefore, a high-precision global synchronization mechanism must be established.

[0111] S23, the original data is stored based on erasure coding.

[0112] A hierarchical storage strategy for hot and cold data is adopted, and erasure coding (EC) technology is applied to cold data to reduce redundant storage overhead.

[0113] For data requiring real-time processing and short-term queries (such as raw event streams from the last few minutes), it is stored in a cache (such as Redis or an in-memory database). This data is used to support real-time situational calculations and anomaly detection within short time windows.

[0114] For trending data that needs to be stored long-term (such as the average hourly entanglement distribution success rate, daily quantum gate error rate changes, etc.), it is periodically transferred from hot data storage to a distributed file system or object storage. Before storage, this data is erasure coded. For example, RS(6,3) encoding is used to divide the original data into 6 data blocks and calculate 3 check blocks, then these 9 blocks are distributed and stored on different storage nodes. The advantage of this technology is that it can fully recover the original data even if any 3 blocks are lost or corrupted, while the storage overhead is only 1.5 times that of the original data (9 / 6), far exceeding the traditional 3-replica strategy (300% storage overhead). This significantly improves data reliability while effectively controlling the cost of long-term storage.

[0115] The purpose of step S23 is to address the issues of reliable storage and high cost of massive long-term monitoring data while ensuring real-time data acquisition. The raw monitoring data (such as the execution results of each quantum gate) is enormous, and storing it all would be prohibitively expensive. However, for trend analysis and retrospective analysis, it is necessary to retain some data with long-term value (such as network utilization and error rate trends).

[0116] The purpose of step S2 is to address the question of where and how monitoring data in a distributed quantum computing network originates. Guided by the state quantity system defined in step S1, it involves deploying physical or virtual probes to reliably and synchronously collect the raw, underlying data required for calculating state quantities from various quantum nodes, network links, and control systems, providing accurate and up-to-date data for the entire monitoring system.

[0117] Step S3: Perform real-time calculation of situational parameters and multi-level fusion on the raw data.

[0118] In this embodiment, step S3 may specifically include the following steps:

[0119] S31, based on a high-performance computing chip, performs real-time parallel computing on the raw data.

[0120] A heterogeneous high-performance computing architecture is adopted to parallelize computing tasks.

[0121] The core utilizes a heterogeneous computing platform combining CPU and GPU or CPU and FPGA. The CPU handles complex control logic and task scheduling, while the GPU or FPGA handles a large number of repetitive, parallel computational tasks. For example, calculating the cumulative quantum gate error rate (CQG). The formula for calculating ) is:

[0122] .

[0123] in, It is the total number of single quantum gates. It is the first Error rate of a single quantum gate; It is the total number of two quantum gates (such as CNOT gates). It is the first Error rate of two quantum gates; It is the total number of measurement operations. It is the first The error rate per measurement operation. This formula involves a large number of decimal multiplications, making it ideal for parallel computation on GPUs. By distributing millions of multiplication operations to thousands of GPU cores for parallel processing, computational latency can be reduced from seconds to milliseconds.

[0124] The entire data processing flow is designed as a multi-stage pipeline. One thread is responsible for receiving and parsing raw data packets from the message queue, another thread (or thread pool) is responsible for performing specific situational quantity calculations, and a third thread is responsible for writing the calculation results to the database and sending them to downstream analysis modules. This pipeline design ensures data throughput and avoids the impact of I / O wait on computational performance.

[0125] The purpose of step S31 is to ensure the real-time calculation of the state parameters, meeting the dynamic control requirements of the DQC system. The state of a quantum system changes rapidly; if the computational delay is too high, the resulting state parameters become historical data and lose their value in guiding current operations.

[0126] S32, based on FPGA, performs continuous quantum error correction in real time.

[0127] A hardware-level error correction control loop is implemented by leveraging the ultra-low latency characteristics of FPGA.

[0128] The FPGA is configured as a dedicated CQEC controller. It is directly connected to the readout line of the auxiliary bits used for parity checking on the quantum chip. The digital logic for parsing the measurement results is implemented internally within the FPGA.

[0129] When the FPGA detects that the measurement result of the auxiliary bit indicates that an error has occurred in a data bit (e.g., a bit flip error), it can generate a correction pulse within tens of nanoseconds and apply it directly to the data bit with the error via a high-speed waveform generator. This process is completed entirely at the hardware level, without needing to report the measurement results to the CPU for software decision-making, thus achieving true real-time error correction. Simultaneously, the FPGA reports the time, type, and correction action of the error as raw data for distributed error correction status metric calculation to the monitoring system for calculating macroscopic indicators such as the logic error rate.

[0130] Step S32 serves to enable real-time monitoring and immediate intervention of distributed error correction status variables (such as the logic error rate). In quantum error-correcting codes, the parity of ancillaqubits needs to be measured frequently (typically on the nanosecond to microsecond scale) to detect errors. This high-frequency measurement and feedback cannot be accomplished in software and must be implemented in a closed loop at the hardware level.

[0131] S33 uses data fusion to perform feature-understanding-evaluation three-layer information fusion.

[0132] Construct a hierarchical data fusion model. Perform feature layer fusion, which directly processes the raw data from step S2. For example, for the abstract concept of node computing power, feature layer fusion will reduce and integrate multiple raw features such as the collected quantum bit T1 / T2 time, single-gate fidelity, two-gate fidelity, and readout fidelity through methods such as normalization and principal component analysis to generate a multi-dimensional feature vector.

[0133] The understanding layer then performs fusion, taking the feature vectors output from the feature layer and the individual state variables (such as dynamic quantum volume and number of available qubits) calculated in the first step as input. Its goal is to understand the relationships between these metrics. For example, using Bayesian networks, the probabilistic dependencies between quantum communication latency and entanglement fidelity, as well as node load, can be learned. When the understanding layer detects a decrease in entanglement fidelity, it can infer from the learned model that this may be due to increased communication latency or exacerbated node crosstalk.

[0134] The evaluation layer then performs fusion, which is the highest level of fusion. It combines the various relationships and identified patterns output from the understanding layer with pre-defined expert rules and system objectives, ultimately outputting one or more comprehensive indices reflecting the overall health of the system. For example, the comprehensive system health index might be a weighted sum of node computing power, network quality, and task completion rate. A typical formula could be: .

[0135] in, , , It is a weighting coefficient that can be dynamically adjusted based on the main task type currently running on the system (whether it is computationally intensive or communication-intensive). This represents the current node's computing power (or the current performance value related to computing power). This represents the node's maximum computing power (or a baseline computing power value). For computing power utilization or computing power health, This refers to the task success rate (which may refer to the percentage of tasks that are successfully completed). This represents the average progress of tasks (e.g., the average completion progress of multiple tasks, reflecting overall progress efficiency). This comprehensive index provides operations and maintenance personnel with a clear overview of the system status.

[0136] The purpose of step S33 is to eliminate the uncertainty and noise caused by a single sensor or single-dimensional data, and to generate a comprehensive judgment on the overall state of the system through data fusion, so as to avoid the blind men and the elephant-like one-sided cognition.

[0137] Step S3 aims to address the processing and transformation of raw data into high-value information. Its function is to take the massive, noisy, and low-level raw data collected in step S2 and, according to the formulas and algorithms defined in step S1, calculate in real time a situational quantity with clear physical meaning. Furthermore, through multi-level data fusion technology, the dispersed situational quantities are integrated into a comprehensive assessment reflecting the overall health of the system, eliminating data redundancy and noise, and providing clear, accurate, and condensed information for higher-level decision-making.

[0138] Step S4: Analyze the intrinsic relationships between the state variables and the patterns of their evolution over time to obtain the analysis results.

[0139] In this embodiment, step S4 may specifically include the following steps:

[0140] S41, based on Bayesian networks, performs situational correlation modeling.

[0141] Construct a structure learning and parameter learning system based on Bayesian networks.

[0142] First, leveraging expert knowledge, preliminary causal assumptions are defined between state variables to construct a prototype of a directed acyclic graph (DAG). For example, it can be assumed that quantum communication latency and entanglement fidelity directly affect the probability of communication interruption, while the probability of communication interruption and the number of available qubits at a node jointly affect the progress of subtask completion.

[0143] Subsequently, using historical monitoring data, the conditional probability table (CPT) of each node in the Bayesian network is trained using the likelihood-weighted method or the expectation-maximization (EM) algorithm, quantifying the dependency weights between nodes. For example, the system can learn the specific probability of a subtask's completion being delayed when the probability of communication interruption is high. After training, this Bayesian network becomes a powerful inference engine. When certain observed situational variables (evidence) are input, it can calculate the posterior probabilities of other unobserved or future situational variables, thereby achieving causal inference.

[0144] The purpose of step S41 is to reveal the intrinsic connections and mutual influence mechanisms between different dimensions of situational variables. For example, how does network congestion (increased communication latency) affect the error rate (cumulative error rate) of task execution, and how does fluctuation in node computing power (decline in dynamic quantum volume) affect the overhead of distributed error correction (the proportion of error correction code overhead)? By constructing a probabilistic graphical model, these complex dependencies can be clearly quantified.

[0145] S42, based on hybrid quantum classical LSTM, is used for time series prediction.

[0146] Construct a hybrid quantum classical long short-term memory (LSTM) neural network model.

[0147] Traditional LSTM struggles to directly process high-dimensional quantum state data. This method first uses a parameterized quantum circuit (PQC) as an embedding layer to map the input time-series data (such as a sequence of quantum gate error rates over a past period) into a higher-dimensional Hilbert space. This quantum embedding can capture quantum correlation features in the data that are difficult for classical models to detect.

[0148] The output of the quantum embedding layer serves as the input to a classic LSTM network. LSTM networks excel at handling time-series data and can learn long-term dependencies within the data. By training on historical data, LSTM networks can learn patterns in error rate evolution; for example, it might learn that "after three consecutive calibrations, the error rate typically exhibits a slow upward trend."

[0149] The final output of the hybrid model is a prediction of the status parameters over a future time window. For example, it can predict the trend of the quantum gate error rate over the next 10 minutes, or whether the progress of a subtask will reach its target value within the next 5 minutes. This prediction will serve as an important input for subsequent risk assessment.

[0150] Step S42 aims to predict the future evolution trends of key state variables. The performance of quantum systems (such as the coherence time of qubits and gate error rate) typically drifts and decays over time, exhibiting complex temporal patterns. Predicting these trends in advance is crucial for proactive maintenance and fault-tolerant scheduling.

[0151] S43, based on Markov decision processes, performs system behavior modeling.

[0152] Mapping the core elements of the DQC system to the quintuples of MDP. middle.

[0153] State space (S): Defines the system's state vector. At each step, the system's state consists of the key status variables at the current moment. For example, ,in It is the first The dynamic quantum volume of each node It is the average communication delay. This is the average progress of all tasks. This state vector It is the system at the current moment A full snapshot.

[0154] Action space (A): Defines the operations that the scheduler can perform. For example, A = {Assign task A to node 1, assign task B to node 2, ..., adjust the error correction code level of task A, ...}.

[0155] State transition probability (P): Defined in state Next action Afterwards, the system transitions to a new state. The probability. This probability can be provided by the Bayesian network and the prediction model in step S4.

[0156] Reward function (R): Defines the immediate benefit of taking an action in each state. For example, the reward may be linked to the number of high-priority tasks completed, the system's average throughput, or penalties for violating a Service Level Agreement (SLA). .

[0157] Discount factor ( ): Used to weigh the importance of immediate rewards against future long-term rewards.

[0158] By constructing this MDP model, a complex dynamic scheduling problem is transformed into finding the optimal policy (i.e., the mapping from state to action) within the MDP. This strategy addresses the issue of maximizing long-term accumulated discount rewards.

[0159] Step S43 aims to abstract the operation and scheduling of the entire DQC system into an optimizable mathematical model. The MDP framework can formally describe how the system state transitions under different states (represented by status variables) and how much immediate or long-term benefit (such as task completion rate) is gained after taking different actions (such as scheduling decisions). This provides a solid theoretical foundation for subsequent intelligent decision-making.

[0160] The purpose of step S4 is to address the problem of gaining a deep understanding of the system's state. After obtaining real-time and accurate situational variables in step S3, it analyzes the inherent relationships between these variables and their evolution over time to gain insight into the essence of system operation. This goes beyond simply knowing what is happening now; it's about understanding why it happened and what will happen next, thereby providing causal and trend-based insights for risk assessment and optimization decisions.

[0161] Step S5: Based on the analysis results, conduct dynamic network risk assessment and graded early warning.

[0162] In this embodiment, step S5 may specifically include the following steps:

[0163] S51. Based on the analysis results, construct a multivariate risk scoring model.

[0164] Construct a comprehensive risk scoring model based on weighted sums or more complex nonlinear functions.

[0165] Based on the risk score of task failure ( For example, this score can integrate the following key predictive indicators:

[0166] Cumulative quantum gate error rate ( ): This is calculated in real time by step S3 and reflects the errors that have accumulated in the task.

[0167] Predicted value of communication interruption probability ( ): This is predicted by the hybrid quantum classical LSTM model in step S4, reflecting the probability that the quantum network may be interrupted within the future task execution window.

[0168] Predicted value of computing power overload risk ( ): Also derived from the prediction model, it reflects the risk that the node where the task is located may experience extended task execution time or failure in the future due to insufficient computing power.

[0169] Scoring Function: A typical risk scoring function can be defined as follows:

[0170] .

[0171] in, , , These are the preset baseline thresholds for cumulative error rate, interruption probability, and overload risk, used for normalization. , , It is a weighting coefficient, and They can be adjusted according to the task type: for a communication-sensitive task (such as quantum teleportation), they can be... Set it higher; for a computationally intensive task (such as Shor's algorithm), it can be set higher. Set it higher. The final one. It is a value between 0 and 1 (or higher). The larger the value, the higher the risk of task failure.

[0172] The purpose of step S51 is to integrate multiple related situational variables with different dimensions into a unified risk score, thereby quantitatively measuring the overall risk level of the system in a certain aspect (such as the risk of mission failure). This avoids the one-sidedness of triggering alarms with a single indicator.

[0173] S52 maps continuous risk scores to warning levels and defines response strategies for different levels.

[0174] Set multi-level dynamic thresholds and associate them with predefined actions.

[0175] Classification Definition: The system classifies risks into three levels based on risk score values, for example:

[0176] Low risk ( The system is running normally; only logs are being recorded for maintenance personnel to review and reference.

[0177] Medium risk ( The system has potential vulnerabilities. This triggers a level-of-concern alert, notifying operations and maintenance personnel to monitor relevant metrics via push notifications (such as WeChat Work or Slack). Simultaneously, the system begins preparing resources for potential interventions, such as pre-allocating backup computing or communication resources.

[0178] High risk ( The system faces an imminent danger. A critical alert is triggered, and maintenance personnel are immediately notified via audible and visual alarms, telephone notifications, etc. Simultaneously, the system automatically initiates the response process in step S5, such as task migration or degradation.

[0179] Dynamic thresholds: To avoid false alarms caused by periodic fluctuations in system load, these tiered thresholds can be dynamically adjusted based on time (such as peak / off-peak periods) or system status. For example, during peak periods, the high-risk threshold can be adjusted from 0.7 to 0.8 to avoid triggering unnecessary emergency responses due to brief periods of high load.

[0180] The purpose of step S52 is to map continuous risk scores into discrete, operationally meaningful warning levels and define the system's response strategies for different levels. This achieves a leap from numerical monitoring to event management, enabling the system to take differentiated actions based on the severity of the risk.

[0181] S53, Generate a standardized risk assessment report.

[0182] Design a standardized report template and automatically populate the content.

[0183] Report Content: A standardized risk assessment report may include the following sections:

[0184] Report Overview: Date of creation, report validity period, scope of the system assessed, and overall risk level (e.g., low / medium / high).

[0185] Key risk indicators: List the main risk scores (such as task failure risk, computing power overload risk), and provide their values, trends (rising / stable / falling) and brief explanations.

[0186] List of Major Risk Items: A detailed list of the top three risk items. For each risk item, the following must be clearly stated:

[0187] Risk description: For example, node QPU-03 has a high risk of computing power overload.

[0188] Risk level: High risk.

[0189] Relevant state parameters: List the key indicators that contribute to this risk, such as the dynamic quantum volume of the node. (Below the warning threshold of 5.0), Number of available qubits (Lower than expected).

[0190] Potential impact: For example, it may cause the execution time of tasks B and C running on this node to exceed the SLA by 1.5 times.

[0191] Recommended measures: For example, it is recommended to migrate tasks B and C to node QPU-05.

[0192] Historical trend chart: Includes a trend chart of key risk indicators over the past 24 hours or 7 days.

[0193] Report distribution: Reports can be automatically generated in PDF or HTML format and sent to subscribers via email or made available to users at any time via a web interface.

[0194] The purpose of step S53 is to present the results of the risk assessment in a structured, clear and easy-to-understand format to users in different roles (such as operations and maintenance personnel, system architects, and application developers), and to archive it as a historical record for post-event analysis and auditing.

[0195] The purpose of step S5 is to transform the understanding and prediction of the system status into a quantitative assessment and warning of potential risks. Following the analysis results of step S4, it not only determines whether a problem has occurred currently, but also predicts potential future crises (such as computing overload, communication interruption, and task failure). Based on the severity and urgency of the risks, it generates standardized assessment reports and tiered warning signals, providing a basis for decision-making regarding manual or automatic intervention of the system.

[0196] Step S6: Based on the analysis results, conduct dynamic network risk assessment and graded early warning.

[0197] In this embodiment, step S6 may specifically include the following steps:

[0198] S61, based on dynamic risk assessment and hierarchical early warning results, performs real-time scheduling optimization through deep reinforcement learning.

[0199] The MDP model established in step S4 is combined with a deep reinforcement learning algorithm (such as DQN or PPO).

[0200] Train a DRL agent to act as a central scheduler. The agent's input is the system state defined in the MDP. (Constituted from key state variables). The agent's output is an action. It can be scheduling decisions such as assigning task X to node Y, adjusting the error correction code level of task X, and reserving bandwidth for the link connecting A and B.

[0201] The agent's goal is to maximize long-term cumulative rewards. .in, This is the discount factor, and its value range is usually [range missing]. Used to balance short-term and long-term rewards. The closer the value is to 1, the more the agent focuses on long-term gains. Let be the time step, representing the first interaction between the agent and the environment. step; For the first Instant rewards earned with each step; The cumulative reward, calculated from the initial time step, is the objective function for optimizing the deep reinforcement learning agent. The reward is determined by both the risk assessment result from step S5 and the direct execution result of the task (e.g., success / failure, execution time). For example, a successful task scheduling can earn a positive reward, while a task failure successfully avoided due to accurate risk prediction can earn a higher positive reward. Conversely, if the scheduling decision leads to task failure or exacerbates the risk, the agent will receive a negative reward. Through millions of interactive trials with the environment (simulated or real DQC system), the DRL agent gradually learns an optimal scheduling strategy. This strategy can make decisions that best balance short-term gains and long-term stability based on the real-time system status.

[0202] The purpose of step S61 is to make optimal task scheduling and resource allocation decisions in real time within a dynamically changing DQC environment to maximize long-term system performance (such as throughput and task success rate) and minimize risk. This is a typical sequential decision problem, which is very suitable for solving using deep reinforcement learning.

[0203] S62 will link graded early warning with intelligent response to achieve automatic closed-loop response to early warning signals.

[0204] Establish an event-driven workflow engine that associates alert events with automated scripts or API calls.

[0205] Predefined strategy linkage: For common risks with clear impact, the system can pre-define atomic response strategies. For example:

[0206] Event: The dynamic quantum volume of node QPU-05 is below the danger threshold.

[0207] Response action: Automatically call the Kubernetes Operator's API to mark the node as unschedulable and trigger a smooth migration of all running non-critical tasks on it.

[0208] Event: The entanglement distribution success rate of the link connecting QPU-02 and QPU-03 remains below the attention threshold.

[0209] Response action: Notify the Q-PATH quantum routing algorithm module to assign a lower path weight to this link within the next 5 minutes, guiding new tasks to avoid this unstable link.

[0210] Dynamic Strategy Coordination: For more complex scenarios, warning events can be input as contextual information to the DRL agent, which then dynamically generates a sequence of actions. For example, when the system detects that the task failure risk score exceeds the high-risk threshold, it transmits this event and the current system state to the DRL agent. After rapid reasoning, the agent may output a complex combination of actions, such as migrating task A from node 1 to node 3, simultaneously adding two auxiliary bits of error correction resources for another task B on node 3, and placing a low-priority analysis task C into the waiting queue. This dynamically generated strategy is more flexible than predefined scripts and better adaptable to complex and changing system environments.

[0211] The purpose of step S62 is to connect the risk assessment results of step S5 with the scheduling optimization capabilities of step S6, thereby achieving an automatic closed-loop response to the early warning signal. When the risk score exceeds a preset threshold, the system can not only issue an alarm but also automatically trigger predefined or real-time response strategies generated by the DRL agent.

[0212] S63 evaluates the effectiveness of implemented decisions and uses newly generated data for continuous training and optimization.

[0213] Build an online learning framework that includes data pipelines, model retraining, and A / B testing.

[0214] After each scheduling decision is made, the system closely monitors its subsequent effects. For example, if the agent decides to migrate task A to node 3, the system records metrics such as the success rate, execution time, and resource consumption of task A on node 3, and compares these metrics with the simulated predictions if task A remained on the original node, forming a decision feedback sample. ).in: The current state refers to the state of the system at time t, before a scheduling decision is made. Real-time status. Includes information such as node computing power, network quality, task queue, and risk indicators. The current action refers to the state in which the agent is in action. The scheduling decisions made are, for example, moving task A to node 3. for Instant reward for performing an action Then, the system will provide a reward value at the next moment. This reward is determined by the task execution result (success / failure, execution time) and the risk assessment result. The next state refers to the action to be performed. Afterwards, the system at time The new state to which it has migrated. These massive amounts of feedback samples are continuously collected and stored in an experience replay pool.

[0215] Periodically (e.g., daily or weekly), use the latest data accumulated in the replay pool to perform asynchronous incremental training or fine-tuning on the Bayesian network of step S4, the hybrid quantum classical LSTM prediction model, and especially the DRL agent of step S6. For example, the conditional probability table in the Bayesian network can be updated with new data to more accurately reflect the actual dependencies of the current hardware; or the DRL agent can be further trained with new data to strengthen decision patterns that yielded high rewards in the past week and weaken decision patterns that led to adverse consequences.

[0216] The updated models will not be directly deployed to replace the old models. They will first be deployed to a shadow mode or A / B testing environment. In shadow mode, the new and old models run in parallel, and the decisions of the new model are only recorded and evaluated, not actually executed. Only when the simulation performance of the new model consistently outperforms the old model will the old model in the production environment be smoothly replaced through blue-green deployment or canary release, ensuring the stability and continuous improvement of system performance.

[0217] The purpose of step S63 is to achieve a closed loop in the entire lifecycle of the monitoring and control system, enabling the system to self-evolve. By evaluating the effects of implemented decisions and using newly generated data to continuously train and optimize various models in the system, the system can adapt to hardware aging, environmental changes, and the evolution of task load patterns.

[0218] Step S6 serves as the execution and evolutionary link in the entire monitoring method. Based on the perception, understanding, prediction, and evaluation results of the previous five steps, it proactively takes action to mitigate risks, optimize performance, and form a closed loop from monitoring to action to feedback. Simultaneously, by recording the effects of each action, the system can continuously learn and optimize its decision-making model, achieving continuous adaptive evolution.

[0219] The beneficial effects of implementing this embodiment are:

[0220] (1) By defining and configuring the situation quantity system, the four dimensions of quantum hardware state, network performance, error correction effect and task progress are uniformly quantitatively modeled, realizing the all-round situational awareness of the DQC system from the physical layer to the task layer, providing a complete data foundation for the understanding and control of complex systems, constructing a multi-dimensional situational awareness system, and breaking through the limitations of single-dimensional evaluation.

[0221] (2) By performing multi-level fusion of multi-source heterogeneous data, it is possible not only to dynamically track changes in system state, but also to explore the intrinsic correlation and evolution law between situational quantities, thereby shifting from passive monitoring to active cognition, providing high-quality real-time intelligence for subsequent decision-making, realizing real-time calculation and deep fusion of situational quantities, and improving the dynamics and accuracy of monitoring.

[0222] (3) Based on situational analysis, a dynamic risk assessment and hierarchical early warning mechanism is introduced, which can identify systemic risks and issue early warnings. More importantly, through the closed-loop process of perception-analysis-assessment-scheduling, adaptive scheduling and continuous optimization of network status are realized, which significantly improves the robustness and execution efficiency of DQC network in dynamic environment, and has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control.

[0223] This invention can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0224] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0225] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0226] Example 2

[0227] Further reference Figure 2 As a response to the above Figure 1 The present invention provides an embodiment of a computing power network status monitoring device, which is similar to the method shown. Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0228] like Figure 2 As shown, the computing power network status monitoring device 70 in this embodiment includes: a definition module 71, a data acquisition module 72, a fusion module 73, an analysis module 74, an evaluation module 75, and an optimization module 76. Wherein:

[0229] Module 71 is defined to define and configure the state quantity system, mapping the quantum hardware state and network behavior into quantitative indicators with clear physical meaning and computational methods;

[0230] The acquisition module 72 is used to acquire the raw data required for calculating the state quantity from various quantum nodes, network links and control systems based on the state quantity system;

[0231] The fusion module 73 is used to perform real-time calculation of situational parameters and multi-level fusion on the original data;

[0232] Analysis module 74 is used to analyze the intrinsic relationships between state variables and their evolution over time to obtain analysis results;

[0233] The assessment module 75 is used to conduct dynamic network risk assessment and graded early warning based on the analysis results.

[0234] Optimization module 76 is used to perform adaptive scheduling and closed-loop optimization of network status based on dynamic risk assessment and hierarchical early warning results.

[0235] The beneficial effects of implementing this embodiment are: it constructs a multi-dimensional situational awareness system, breaking through the limitations of single-dimensional assessment; it realizes real-time calculation and deep fusion of situational quantities, improving the dynamism and accuracy of monitoring; and it has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control.

[0236] Example 3

[0237] Further reference Figure 3 As a response to the above Figure 1 The present invention provides another embodiment of a computing network status monitoring device based on the method shown, which is similar to... Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0238] like Figure 3 As shown, the computing power network status monitoring device described in this embodiment includes: a data acquisition module, a data processing module, a correlation analysis module, and a risk assessment and scheduling module. The data acquisition module is responsible for collecting multiple core status variables in real time from multi-node quantum hardware and the quantum network. The data processing module is used to preprocess, correct, and transform the collected status variable data. The correlation analysis module is used to analyze the correlation relationships between status variables based on Bayesian networks and a hybrid quantum classical LSTM model. The risk assessment and scheduling module is used to predict system risks and dynamically adjust task scheduling strategies based on the correlation analysis results. The data acquisition module adopts distributed acquisition, low-latency synchronization, and high-reliability storage to ensure the real-time performance and integrity of the monitoring data. The data processing module uses a status variable calculation and fusion method to transform underlying data into quantified status variables and comprehensive evaluation results; the core is real-time calculation and multi-level fusion. The correlation analysis module is used to reveal the inherent laws of the system by constructing a multi-dimensional status variable correlation model, realizing real-time monitoring, trend prediction, and risk assessment, providing decision support for subsequent early warning and optimization. The risk assessment and scheduling module, based on the correlation analysis results, realizes system risk prediction and dynamic scheduling.

[0239] The beneficial effects of implementing this embodiment are: it constructs a multi-dimensional situational awareness system, breaking through the limitations of single-dimensional assessment; it realizes real-time calculation and deep fusion of situational quantities, improving the dynamism and accuracy of monitoring; and it has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control.

[0240] Example 4

[0241] To address the aforementioned technical problems, embodiments of the present invention also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0242] The aforementioned computer device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected via a system bus. It should be noted that only the computer device 8 with components 81, 82, and 83 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described herein is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0243] The aforementioned computer devices can be desktop computers, laptops, handheld computers, and cloud servers, among other computing devices. These devices can facilitate human-computer interaction with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0244] The aforementioned memory 81 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the aforementioned memory 81 may be an internal storage unit of the aforementioned computer device 8, such as the hard disk or memory of the computer device 8. In other embodiments, the aforementioned memory 81 may also be an external storage device of the aforementioned computer device 8, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 8. Of course, the aforementioned memory 81 may also include both the internal storage unit and its external storage device of the aforementioned computer device 8. In this embodiment, the aforementioned memory 81 is typically used to store the operating system and various application software installed on the aforementioned computer device 8, such as computer-readable instructions for a computing network status monitoring method. In addition, the aforementioned memory 81 can also be used to temporarily store various types of data that have been output or will be output.

[0245] In some embodiments, the processor 82 described above may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 82 is typically used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to execute computer-readable instructions stored in the memory 81 or to process data, for example, to execute computer-readable instructions for the computing network status monitoring method described above.

[0246] The network interface 83 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the computer device 8 and other electronic devices.

[0247] The beneficial effects of implementing this embodiment are: it constructs a multi-dimensional situational awareness system, breaking through the limitations of single-dimensional assessment; it realizes real-time calculation and deep fusion of situational quantities, improving the dynamism and accuracy of monitoring; and it has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control.

[0248] Example 5

[0249] The present invention also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the computing power network status monitoring method described above.

[0250] The beneficial effects of implementing this embodiment are: it constructs a multi-dimensional situational awareness system, breaking through the limitations of single-dimensional assessment; it realizes real-time calculation and deep fusion of situational quantities, improving the dynamism and accuracy of monitoring; and it has forward-looking risk assessment and closed-loop optimization capabilities, making up for the shortcomings of existing systems in prediction and control.

[0251] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0252] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for monitoring the status of a computing network, characterized in that, Includes the following steps: Define and configure the state quantity system, which maps the quantum hardware state and network behavior into quantitative indicators with clear physical meaning and calculation methods. The state quantity includes core state quantities of node computing power dimension, quantum network dimension, distributed error correction dimension and task execution dimension. Based on the aforementioned state quantity system, raw data required for calculating the state quantity are collected from various quantum nodes, network links, and control systems; The original data is used to calculate the situational quantity in real time, and multi-level fusion is performed based on a three-layer information fusion method of feature layer fusion, understanding layer fusion and evaluation layer fusion. The intrinsic correlation between state variables is analyzed based on Bayesian network analysis, and the evolution of state variables over time is predicted by time series based on hybrid quantum classical long short-term memory neural network analysis to obtain the analysis results. Based on the analysis results, a multivariate risk scoring model is constructed to map continuous risk scores to early warning levels, thereby conducting dynamic network risk assessment and graded early warning. Based on dynamic risk assessment and hierarchical early warning results, real-time scheduling optimization is performed through deep reinforcement learning, hierarchical early warning and intelligent response are linked and executed, and the network state adaptive scheduling and closed-loop optimization are achieved by evaluating the effect of the executed decisions and using newly generated data for continuous training and optimization.

2. The computing network status monitoring method according to claim 1, characterized in that, The defined and configured state quantity system maps quantum hardware states and network behaviors into quantitative indicators with clear physical meaning and computational methods. The specific steps for defining and configuring core state quantities, including those representing node computing power, quantum network dimensions, distributed error correction dimensions, and task execution dimensions, include: Standardize the definitions of core state variables in the dimensions of node computing power, quantum network, distributed error correction, and task execution. Configure differentiated correction formulas for different hardware routes for core status variables; Multiple levels of early warning thresholds are preset for core situational variables.

3. The computing network status monitoring method according to claim 1, characterized in that, The steps for collecting the raw data required for calculating the state quantity from various quantum nodes, network links, and control systems based on the state quantity system specifically include: Based on the aforementioned situational quantity system, a distributed acquisition terminal is deployed to achieve standardized encapsulation of hardware interfaces and collect the raw data required for calculating situational quantities. Based on gRPC, the collected raw data is synchronized with low latency. The original data is stored based on erasure coding.

4. The computing network status monitoring method according to claim 1, characterized in that, The steps of performing real-time situational quantity calculations on the raw data and conducting multi-level fusion based on a three-layer information fusion approach of feature layer fusion, understanding layer fusion, and evaluation layer fusion specifically include: Based on a high-performance computing chip, the raw data is subjected to real-time parallel computation. Real-time feedback of continuous quantum error correction based on FPGA; By using data fusion, we can achieve three-layer information fusion: feature-understanding-evaluation.

5. The computing network status monitoring method according to claim 1, characterized in that, The steps for obtaining the analysis results, specifically, include: analyzing the intrinsic correlations between state variables based on Bayesian networks and performing time series predictions on the evolution of state variables over time based on hybrid quantum classical long short-term memory neural networks. Based on Bayesian networks, we perform situational correlation modeling. Time series prediction based on hybrid quantum classical LSTM; System behavior modeling is performed based on Markov decision processes.

6. The computing network status monitoring method according to claim 1, characterized in that, The steps of constructing a multivariate risk scoring model based on the analysis results, mapping continuous risk scores to early warning levels, and conducting dynamic network risk assessment and graded early warning specifically include: Based on the analysis results, a multivariate risk scoring model is constructed; Map continuous risk scores to warning levels and define response strategies for different levels; Generate a standardized risk assessment report.

7. The method for monitoring the status of a computing network according to any one of claims 1 to 6, characterized in that, The steps of achieving network state adaptive scheduling and closed-loop optimization based on dynamic risk assessment and hierarchical early warning results, using deep reinforcement learning for real-time scheduling optimization, linking hierarchical early warning with intelligent response, evaluating the effectiveness of executed decisions and using newly generated data for continuous training and optimization, specifically include: Based on dynamic risk assessment and hierarchical early warning results, real-time scheduling optimization is performed through deep reinforcement learning; By linking tiered early warning with intelligent response, an automatic closed-loop response to early warning signals can be achieved. By evaluating the effectiveness of implemented decisions and using newly generated data for continuous training and optimization.

8. A computing network status monitoring device, characterized in that, include: The definition module is used to define and configure the state quantity system, mapping the quantum hardware state and network behavior into quantitative indicators with clear physical meaning and calculation methods. The state quantities include core state quantities of node computing power dimension, quantum network dimension, distributed error correction dimension and task execution dimension. The acquisition module is used to acquire the raw data required for calculating the state variables from various quantum nodes, network links, and control systems based on the state variable system. The fusion module is used to perform real-time calculation of situational parameters on the raw data and to perform multi-level fusion based on a three-layer information fusion method of feature layer fusion, understanding layer fusion and evaluation layer fusion. The analysis module is used to analyze the intrinsic correlation between state variables based on Bayesian networks, and to perform time series prediction on the evolution of state variables over time based on hybrid quantum classical long short-term memory neural networks, thereby obtaining analysis results. The assessment module is used to construct a multivariate risk scoring model based on the analysis results, map continuous risk scores to early warning levels, and conduct dynamic network risk assessment and graded early warning. The optimization module is used to perform real-time scheduling optimization based on dynamic risk assessment and hierarchical early warning results through deep reinforcement learning, link hierarchical early warning with intelligent response, and continuously train and optimize the network state by evaluating the effect of the executed decisions and using newly generated data, so as to achieve adaptive scheduling and closed-loop optimization of the network state.

9. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the computing power network status monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the computing power network status monitoring method as described in any one of claims 1 to 7.

Images