Search

CN-122019175-A - Computer resource intelligent allocation method and system based on artificial intelligence

CN122019175ACN 122019175 ACN122019175 ACN 122019175ACN-122019175-A

Abstract

The invention discloses a computer resource intelligent distribution method and system based on artificial intelligence, which belongs to the technical field of artificial intelligence, and comprises the steps of constructing a cross-level resource dependency graph by collecting multi-layer resource monitoring data, combining root cause sensitivity with historical time sequence data by fusing causal contribution degree and time dimension dynamic characteristics, accurately predicting resource demand fluctuation trend by LSTM, and the time elastic coefficient of the task to the resource is quantized, and a resource time elastic matrix is generated, so that the resource allocation has dynamic time sequence adaptive capacity, resource idling and sudden congestion are effectively avoided, the resource multiplexing rate and the task scheduling flexibility are improved, the elastic scheduling space is released while the key service SLA is ensured, and the toughness and the resource utilization efficiency of the system for coping with load fluctuation are obviously enhanced.

Inventors

  • ZHOU TAOYI

Assignees

  • 杭州万向职业技术学院

Dates

Publication Date
20260512
Application Date
20260203

Claims (10)

  1. 1. The intelligent computer resource allocation method based on artificial intelligence is characterized by comprising the following steps: s1, constructing a cross-level resource dependency graph by collecting multi-layer resource monitoring data; s2, according to the resource dependency graph, performing root cause positioning of resource conflict through causal discovery, and outputting a causal network describing root causes; s3, based on a causal network of a dependency graph and an elastic modeling, performing plasticity optimization on the time dimension of the resource demand through time sequence prediction and the elastic modeling, and outputting a resource-time elastic matrix; s4, quantifying the coupling effect between heterogeneous resources through nonlinear dynamics analysis based on the dependency graph and the elastic model, and outputting a resource coupling function; s5, carrying out distributed resource negotiation through multi-agent game according to the elastic matrix and the coupling function, and outputting an allocation strategy under Nash equilibrium; S6, carrying out self-adaptive reconstruction of physical resource topology according to an allocation strategy, and outputting an optimized system resource allocation state; S7, performing iterative optimization of the causal distribution model by comparing the inverse fact reasoning with the model according to actual execution feedback of the reconstructed system, and outputting updated model parameters.
  2. 2. The method for intelligently distributing computer resources based on artificial intelligence according to claim 1, wherein in S1, structured monitoring data are collected in real time from a hardware layer, an operating system layer and an application layer, time stamp alignment is carried out on the collected data based on a unified time window, time sequence deviation of high-frequency data of the hardware layer and low-frequency data of the application layer is eliminated, resource indexes of different dimensions are mapped to a [0,1] standardized space through dynamic normalization processing, dimensional differences are eliminated, standardized multi-layer resource state vectors are mapped to a high-dimensional semantic space based on a pre-trained time sequence embedding model, time dynamic and hierarchical relevance are reserved at the same time for the generated embedding vectors, propagation weights among resource nodes are calculated through a graph neural network based on the embedding vectors, an interlayer dependency strength matrix is output, and a dependency graph G (V, E, W) is constructed based on the propagation strength matrix.
  3. 3. The method for intelligently allocating computer resources based on artificial intelligence according to claim 2, wherein in S2, all node pairs are extracted from the resource dependency graph G (V, E, W) constructed in S1 Weighting of As an edge set of initial causal hypothesis, selecting time sequence data of resource conflict events, and aiming at each edge in the dependency graph Performing a conditional independence test, edge The conditional mutual information value of (1) is realized as: , In the formula (i) the formula (ii), Representing edges Is a conditional mutual information value of (1), Representing nodes under condition C Is defined, H represents the entropy function, Representing the source resource node(s), Representing the node of the target resource, Representing a system performance node.
  4. 4. The intelligent distribution method for computer resources based on artificial intelligence according to claim 3, wherein in S2, all triples are traversed for the processed dependency graph When (when) And (3) with Not adjacent and at a given position Under the condition of (a) and (b), And (3) with Determining collision structure when not independently established Based on the determined collision structure, carrying out directional propagation on the remaining undirected edges through Meek rules, and outputting a directed acyclic graph; converting the directed acyclic graph into a causal Bayesian network with probability, and realizing the causal Bayesian network with probability as follows: , In the formula (i) the formula (ii), Represents the causal contribution of node X, Representing the set of all paths from node X to system performance node C, Representing the original causal probability of edge e, And (5) representing a preset threshold value, generating a final causal network, and labeling root cause nodes through causal contribution degree sequencing.
  5. 5. The method for intelligent allocation of computer resources based on artificial intelligence of claim 4, wherein S3, root cause nodes and causal contributions are extracted from a causal network, a time sensitive feature is constructed in combination with a system performance index, resource states are aggregated from multi-layer monitoring data according to time windows, and a structured time sequence dataset is generated 。
  6. 6. The method for intelligent allocation of computer resources based on artificial intelligence according to claim 5, wherein in S3, the pre-trained LSTM model predicts the resource demand distribution of the future time window based on the time series T Calculating the time elasticity coefficient of each task t to the resource v by combining the prediction result and the time sensitivity characteristic The realization is as follows: , In the formula (i) the formula (ii), Representing a causal contribution function, Representing the rate of fluctuation of the demand for resources, The function of the standard deviation is represented, Representing a sequence of demand changes, integrating the elastic coefficients into a resource-time elastic matrix And outputting a resource-time elastic matrix A.
  7. 7. The method for intelligently distributing computer resources based on artificial intelligence according to claim 1, wherein in S4, a dependency graph G (V, E, W), a resource-time elastic matrix A and continuously collected multi-layer monitoring historical data are obtained, and for each time point t, the monitoring data are distributed according to resource nodes Alignment, generating a base state For node v, calculate the weighted elastic state Splicing all node states according to time sequence to form a system-level dynamic state sequence ; Based on D as training data, minimizing the predicted state And true state Is used for the mean square error of (c), Outputting a dynamic evolution function with training completion after convergence 。
  8. 8. The method for intelligently allocating computer resources based on artificial intelligence according to claim 7, wherein in S4, the dynamic evolution function is based And a state sequence D for each pair of resource nodes Extracting state subsequences from D By means of Generating an enhanced track and calculating mutual information Outputting the mutual information matrix ; Calculating average partial derivatives of the comparable matrix on the track according to the mutual information matrix and the dynamic evolution function ; The resource coupling function is: , In the formula (i) the formula (ii), A system state vector is represented and is used to represent, Representing a sign function.
  9. 9. The method for intelligently allocating computer resources based on artificial intelligence according to claim 1, wherein in S5, a resource-time elastic matrix, a resource coupling function, a current task queue and a real-time resource pool state are obtained, each task to be allocated is instantiated as an independent agent, and a physical resource pool is modeled as a shared environment; based on the intelligent agent and the environment model, broadcasting the resource demand intention for the first time, carrying out local adjustment for the second time based on the environment feedback, triggering conflict early warning by the environment when a plurality of intelligent agent requests have antagonistic effect resource combination, guiding the intelligent agent to actively adjust the request; Dynamically calling a coupling function in training to evaluate the quality of resource combination of each round of allocation schemes, taking the effect intensity as a rewarding scaling factor, starting virtual game according to a trained strategy network, a current task and a resource state, calculating an optimal response strategy of each agent based on historical observed opponent strategy distribution, submitting a resource request scheme by each round of iteration, checking a global resource combination effect by using the environment calling coupling function, triggering local strategy fine adjustment if a strong antagonism area is detected, continuously presetting times, ensuring that the strategy fluctuation rate is lower than a threshold value and the global utility fluctuation is stable, mapping an equilibrium strategy into a specific allocation instruction, generating an annotated allocation scheme, marking key synergy effect and evading antagonism points, and outputting the resource allocation strategy under the equilibrium according to the generated allocation scheme.
  10. 10. The computer resource intelligent distribution system based on artificial intelligence realized by the method according to claim 1 comprises a resource dependency graph construction module, a causal reasoning and root cause positioning module, a time sequence elastic modeling module, a resource coupling effect quantization module, a distributed game negotiation module, a topology self-adaptive reconstruction module and a causal model iteration optimization module.

Description

Computer resource intelligent allocation method and system based on artificial intelligence Technical Field The invention belongs to the technical field of artificial intelligence, and particularly relates to an intelligent computer resource allocation method and system based on artificial intelligence. Background With the wide application of artificial intelligence, cloud computing, micro-service architecture and large-scale distributed system, modern computing environments exhibit the characteristics of high isomerism, dynamic coupling and cross-layer dependence, a complex interaction network is formed among hardware resources, operating system resources and application layer services, resource bottlenecks or anomalies at any layer can be conducted to other layers through implicit paths, system-level performance degradation and even service interruption are caused, and the traditional static resource allocation strategy has difficulty in meeting the operation and maintenance requirements of high availability, high efficiency and high elasticity. However, the existing computer resource intelligent allocation method has certain defects, the prior art has systematic defects of rough data fusion, causal reasoning deficiency, insufficient time elastic modeling, cognition on one side of resource coupling effect, lack of synergy and interpretability of allocation mechanisms, weak reconstruction execution safety, incapability of self-evolution of models and the like, so that the method is difficult to accurately perceive cross-layer dependence, accurately position root cause, flexibly schedule time sequence resources and avoid implicit performance conflict in a complex dynamic environment, and finally causes the disconnection of allocation strategies and actual operation, and the resource utilization rate, the system stability and the long-term adaptability are limited. Disclosure of Invention The invention aims to provide an intelligent distribution method and system for computer resources based on artificial intelligence, so as to solve the problems in the background technology. In order to achieve the above purpose, the invention provides the following technical scheme that the computer resource intelligent allocation method based on artificial intelligence comprises the following steps: s1, constructing a cross-level resource dependency graph by collecting multi-layer resource monitoring data; s2, according to the resource dependency graph, performing root cause positioning of resource conflict through causal discovery, and outputting a causal network describing root causes; s3, based on a causal network of a dependency graph and an elastic modeling, performing plasticity optimization on the time dimension of the resource demand through time sequence prediction and the elastic modeling, and outputting a resource-time elastic matrix; s4, quantifying the coupling effect between heterogeneous resources through nonlinear dynamics analysis based on the dependency graph and the elastic model, and outputting a resource coupling function; s5, carrying out distributed resource negotiation through multi-agent game according to the elastic matrix and the coupling function, and outputting an allocation strategy under Nash equilibrium; S6, carrying out self-adaptive reconstruction of physical resource topology according to an allocation strategy, and outputting an optimized system resource allocation state; S7, performing iterative optimization of the causal distribution model by comparing the inverse fact reasoning with the model according to actual execution feedback of the reconstructed system, and outputting updated model parameters. Preferably, in the step S1, structured monitoring data are collected in real time from a hardware layer, an operating system layer and an application layer, time stamp alignment is performed on the collected data based on a unified time window, time sequence deviation of high-frequency data of the hardware layer and low-frequency data of the application layer is eliminated, resource indexes of different dimensions are mapped to a [0,1] standardized space through dynamic normalization processing, dimensional differences are eliminated, a standardized multi-layer resource state vector is mapped to a high-dimensional semantic space based on a pre-trained time sequence embedding model, time dynamics and hierarchy relevance are reserved at the same time for the generated embedding vector, propagation weights among resource nodes are calculated through a graph neural network iteration based on the embedding vector, an interlayer dependent intensity matrix is output, and a dependent graph G (V, E, W) is constructed based on the propagation intensity matrix. Preferably, in the step S2, all node pairs are extracted from the resource dependency graph G (V, E, W) constructed in the step S1Weighting ofAs an edge set of initial causal hypothesis, selecting time sequence data of resource conflict event