An electronic information dynamic scheduling and management system and method

Abstract

The present invention relates to the field of information management technology, and in particular to a dynamic scheduling and management system and method for electronic information. A system data acquisition module collects real-time electronic information data, constructs a resource matrix based on the real-time electronic information data, a task instruction receiving module receives a system task request, parses the system task request to generate a task description vector; a coarse-grained matching module traverses the initial information matrix data based on the task description vector based on an improved Hungarian algorithm to perform coarse-grained matching; a fine-grained optimization module uses the DRL-OPT dynamic reward and punishment reinforcement learning model to perform fine-grained optimization on the candidate node subset data, selects the node that maximizes the long-term reward through Q-Learning, and outputs the optimal node; a system scheduling management module schedules the system according to the resource allocation instruction, can adaptively learn the system operation rules, and dynamically adjust the strategy according to the historical allocation effect.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic information management, and in particular to an electronic information dynamic scheduling and management system and method. Background Art

[0002] Existing electronic information resource scheduling methods are often based on static rules or single metrics, which suffer from the following drawbacks: poor dynamic adaptability, unable to respond in real time to network fluctuations, node load changes, or sudden task demands; low resource coordination efficiency: heterogeneous resources are difficult to uniformly schedule, leading to local resource overload or idleness; high task execution latency, traditional algorithms are prone to falling into local optimality, and global resource matching efficiency is insufficient. Existing technical solutions include task allocation methods based on load balancing, but these fail to consider the coupling relationship between cross-layer resources. Traditional solutions, such as reinforcement learning scheduling models, lack multidimensional feedback mechanisms, resulting in slow convergence in dynamic scenarios. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an electronic information dynamic scheduling and management system and method.

[0004] The technical solution of the present invention to achieve the above object is that, further, in the above-mentioned electronic information dynamic scheduling and management system, the electronic information dynamic scheduling and management system includes the following modules:

[0005] A system data acquisition module is used to deploy a monitoring agent at each physical node, collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data;

[0006] A task instruction receiving module is used to receive a system task request, parse the system task request to generate a task description vector, and calculate an urgency factor to obtain task requirement data;

[0007] A coarse-grained matching module is used to traverse the initial information matrix data to perform coarse-grained matching based on the improved Hungarian algorithm according to the task description vector to obtain candidate node subset data;

[0008] A fine-grained optimization module is used to perform fine-grained optimization on the candidate node subset data using the DRL-OPT dynamic reward and punishment reinforcement learning model, select the node that maximizes the long-term reward through Q-Learning, output the optimal node, and obtain resource allocation instructions;

[0009] The system scheduling management module is used to schedule the system according to the resource allocation instruction, monitor the load of the optimal node, and start the resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

[0010] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the system data acquisition module includes the following submodules:

[0011] The data collection submodule is used to collect the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; the computing resources include CPU utilization, GPU CUDA core occupancy rate and the number of idle FPGA computing units; the storage resources include disk IOPS, remaining storage space and SSD life wear value; the network resources include inter-node delay, remaining bandwidth and TCP retransmission rate; the energy consumption data includes current power and energy consumption per unit computing power;

[0012] A matrix establishment submodule, configured to establish a resource matrix in node order based on the 4-dimensional index of each physical node in the real-time electronic information data;

[0013] The matrix normalization submodule is used to normalize the resources using the minimum-maximum normalization method, so that the indicators of each dimension are within the same value range, and obtain initial information matrix data.

[0014] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the task instruction receiving module includes the following submodules:

[0015] The task receiving submodule is used to receive a system task request, obtain the computational load, memory size, data volume, and deadline in the system task request, and obtain the task computation parameters;

[0016] The urgency calculation submodule is used to integrate the task calculation parameters into a task description vector, obtain the current time and the deadline in the task description vector, and calculate the urgency factor of the task based on the pre-set weight coefficient and service quality level. ;

[0017] ;

[0018] in, and represents the pre-set weight coefficient, Indicates the current time, Indicates the deadline, Indicates the service quality level.

[0019] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the coarse-grained matching module includes the following submodules:

[0020] A traversal submodule is used to traverse the initial information matrix data according to the computational memory requirements and data volume in the task description vector, screen out nodes that meet the basic requirements of the task for computational volume, memory, and data volume, and form the nodes into a first candidate node subset;

[0021] a sorting submodule, configured to sort the first candidate node subset from most to least according to the amount of remaining resources of the nodes, to obtain a second candidate node subset;

[0022] The computation module is used to build a bipartite graph based on the improved Hungarian algorithm, with the task description vector on the left and the second candidate node subset on the right.

[0023] Introducing slack variables for multi-resource constraints , establish the Lagrangian function:

[0024] ;

[0025] in, Representation node The remaining resources, Indicates the matching deviation between node resource efficiency and task description vector; Indicates the resources required by the task description vector;

[0026] A submodule is obtained, which is used to output the Top-5 candidate nodes through row and column reduction and augmented path search, and obtain the candidate node subset data.

[0027] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the fine-grained optimization module includes the following units:

[0028] Model definition unit, used to define the state space, action space and reward function of the DRL-OPT dynamic reward and punishment reinforcement learning model;

[0029] a node output unit, configured to take the real-time state of the candidate node subset data as input as the state space of the model, generate a binary action of selection or non-selection for each node in the candidate node subset using the action space, and output a single optimal node;

[0030] The reward function unit is used to determine that the reward function includes a multi-dimensional dynamic reward and punishment mechanism, including at least positive rewards, negative rewards and dynamic adjustments.

[0031] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the fine-grained optimization module further includes the following units:

[0032] An action selection unit, configured to select a node that maximizes a long-term reward through Q-Learning, collect state data of candidate nodes based on the candidate node subset data, and select an action through an ε-greedy strategy;

[0033] A greedy strategy unit, configured to implement a greedy strategy that at least includes randomly exploring under-evaluated nodes with a probability of ε and selecting a node with the largest current Q value with a probability of 1-ε;

[0034] The reward calculation unit is used to calculate the immediate reward based on the actual results, update the Q value using the Q-Learning formula, output the optimal node, and obtain the resource allocation instruction;

[0035] ;

[0036] in, represents the state-action value function, represents the learning rate, represents the discount factor, Indicates the next state, Indicates the next action. Indicates that the status The maximum Q value that can be obtained from all possible actions.

[0037] Furthermore, in the above-mentioned electronic information dynamic scheduling and management system, the system scheduling management module includes the following units:

[0038] The node management unit is used to encapsulate the optimal node ID and task parameters into a scheduling instruction and send it to the node manager. The node manager verifies the resource availability and starts the task execution if the verification is successful. Otherwise, an exception handling is triggered.

[0039] A threshold setting unit is used to preset multiple thresholds, including at least a warning threshold and an emergency threshold. The warning threshold is ≥80% CPU usage and ≥90% memory usage for 5 minutes; the emergency threshold is ≥95% CPU usage and ≥98% memory usage for 1 minute.

[0040] The resource borrowing and returning unit is configured to start a resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

[0041] Furthermore, in the above-mentioned method for dynamic scheduling and management of electronic information, the method for dynamic scheduling and management of electronic information includes the following steps:

[0042] Deploy a monitoring agent at each physical node to collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data;

[0043] Receiving a system task request, parsing the system task request to generate a task description vector, and calculating an urgency factor to obtain task requirement data;

[0044] Based on the improved Hungarian algorithm, the initial information matrix data is traversed to perform coarse-grained matching according to the task description vector to obtain candidate node subset data;

[0045] The DRL-OPT dynamic reward and punishment reinforcement learning model is used to perform fine-grained optimization on the candidate node subset data, and the node that maximizes the long-term reward is selected through Q-Learning. The optimal node is output and the resource allocation instruction is obtained;

[0046] The system is scheduled according to the resource allocation instruction, and the load of the optimal node is monitored. When the real-time load of the optimal node exceeds a preset threshold, the resource borrowing and returning protocol is started.

[0047] Furthermore, in the above-mentioned electronic information dynamic scheduling and management method, the monitoring agent is deployed at each physical node to collect real-time electronic information data, a resource matrix is ​​constructed based on the real-time electronic information data, and normalized to obtain initial information matrix data, including:

[0048] Collect the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; the computing resources include CPU utilization, GPU CUDA core occupancy rate and the number of idle FPGA computing units; the storage resources include disk IOPS, remaining storage space and SSD life wear value; the network resources include inter-node delay, bandwidth remaining amount and TCP retransmission rate; the energy consumption data includes current power and energy consumption per unit computing power;

[0049] Establishing a resource matrix according to the order of nodes based on the 4-dimensional index of each physical node in the real-time electronic information data;

[0050] The resources are normalized using the minimum-maximum normalization method so that the indicators of each dimension are within the same value range to obtain the initial information matrix data.

[0051] Furthermore, in the above-mentioned electronic information dynamic scheduling and management method, the receiving of a system task request, parsing the system task request to generate a task description vector, and calculating an urgency factor to obtain task requirement data includes:

[0052] Receive a system task request, obtain the computational load, memory size, data volume, and deadline in the system task request, and obtain task computation parameters;

[0053] Integrate the task calculation parameters into a task description vector, obtain the current time and the deadline in the task description vector, and calculate the urgency factor of the task based on the pre-set weight coefficient and the service quality level ;

[0054] ;

[0055] in, and represents the pre-set weight coefficient, Indicates the current time, Indicates the deadline, Indicates the service quality level.

[0056] Its beneficial effects are: 1. By deploying monitoring agents at physical nodes to collect real-time electronic information data and constructing a resource matrix for normalization processing, it ensures that the basic data obtained by the system has high timeliness and standardized characteristics. The resource status information can truly reflect the node capabilities, provide a reliable basis for subsequent task matching, and avoid allocation deviations caused by data lag or format confusion. 2. Intelligent parsing and priority determination of task requirements parses the system task request to generate a task description vector, realizing the quantitative expression of task requirements. Dynamically distinguish between the urgency of tasks and the difference in resource requirements, give priority to guaranteeing the resource supply of high-priority tasks, and effectively improve the overall response efficiency and service quality of the system. 3. Efficient resource screening for coarse-grained matching Based on the improved Hungarian algorithm to traverse the initial information matrix data for coarse-grained matching, it can quickly screen out a subset of candidate nodes that meet the basic conditions of the task from a large number of nodes. Compared with traditional global search, this algorithm significantly reduces the computational complexity and shortens the task allocation decision time. 4. Fine-grained optimized dynamic resource adaptation utilizes the DRL-OPT dynamic reward and punishment reinforcement learning model combined with the Q-Learning algorithm to deeply optimize a subset of candidate nodes, selecting the optimal node with the goal of maximizing long-term rewards. It can adaptively learn the operating rules of the system and dynamically adjust the strategy based on historical allocation results, balancing resource utilization and task completion quality, avoiding local optimal traps, and achieving continuous evolution of resource allocation strategies. 5. Load balancing and elastic resource scheduling build a closed-loop dynamic resource management system by monitoring the load of the optimal node and initiating a resource borrowing and returning protocol. When the node load exceeds the threshold, the borrowing and returning protocol can automatically trigger resource migration or sharing, effectively preventing performance bottlenecks caused by single-point overload and ensuring system stability. At the same time, this mechanism supports elastic scaling of resources, reduces hardware redundancy costs, and improves resource economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0058] Figure 1 This is a schematic diagram of a first embodiment of an electronic information dynamic scheduling and management system according to an embodiment of the present invention;

[0059] Figure 2 Schematic diagram of a second embodiment of a system for dynamic scheduling and management of electronic information according to an embodiment of the present invention;

[0060] Figure 3 Schematic diagram of a third embodiment of a system for dynamic scheduling and management of electronic information according to an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of a first embodiment of a method for dynamic scheduling and management of electronic information in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] Those skilled in the art will understand that, unless otherwise stated, the singular forms "a", "an", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0064] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, a dynamic scheduling and management system for electronic information includes the following modules:

[0065] 101. A system data acquisition module is used to deploy a monitoring agent at each physical node, collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data;

[0066] Specifically, this embodiment also includes a data collection submodule for collecting the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; computing resources include CPU utilization, GPU CUDA core occupancy and the number of idle FPGA computing units; storage resources include disk IOPS, remaining storage space and SSD life wear value; network resources include inter-node delay, bandwidth remaining amount and TCP retransmission rate; energy consumption data includes current power and energy consumption per unit computing power;

[0067] The matrix establishment submodule is used to establish a resource matrix in accordance with the node order based on the 4-dimensional indicators of each physical node in the real-time electronic information data;

[0068] The matrix normalization submodule is used to normalize the resources using the minimum-maximum normalization method, so that the indicators of each dimension are within the same value range and obtain the initial information matrix data.

[0069] 102. A task instruction receiving module is used to receive a system task request, parse the system task request to generate a task description vector, and calculate an urgency factor to obtain task requirement data;

[0070] Specifically, this embodiment also includes a task receiving submodule for receiving a system task request, obtaining the computational load, memory size, data volume, and deadline in the system task request, and obtaining task computation parameters;

[0071] The urgency calculation submodule is used to integrate the task calculation parameters into a task description vector, obtain the current time and deadline in the task description vector, and calculate the urgency factor of the task based on the pre-set weight coefficient and service quality level. ;

[0072] ;

[0073] in, and represents the pre-set weight coefficient, Indicates the current time, Indicates the deadline, Indicates the service quality level.

[0074] Urgency Factor It is a dynamic value calculated by a formula and is used to comprehensively evaluate the two core dimensions of a task: 1. Static service quality requirements, that is, the service quality level, which reflects the importance level of the task; 2. Dynamic time urgency, the difference between the current time and the deadline, which reflects the remaining processing time window through the preset weights. and The linear combination of , ultimately generates the comprehensive urgency score of the task.

[0075] Urgency Factor The core function of the system is to establish a multi-dimensional task priority evaluation system; 1. Intelligent priority sorting, when processing multiple tasks at the same time, the system The value automatically generates the execution queue; 2. Dynamic resource allocation, high Task triggers flexible resource allocation mechanism, low The task is temporarily stored or downgraded; 3. Time limit guarantee mechanism: when the difference between the current time and the deadline approaches zero, the time item will generate exponential growth pressure and automatically trigger the "deadline" emergency processing mode.

[0076] 103. A coarse-grained matching module is used to perform coarse-grained matching based on the task description vector by traversing the initial information matrix data to obtain candidate node subset data;

[0077] Specifically, this embodiment also includes a traversal submodule for traversing the initial information matrix data according to the computational memory requirements and data volume in the task description vector, screening out nodes that meet the basic computational, memory, and data volume requirements of the task, and forming the nodes into a first candidate node subset;

[0078] A sorting submodule, configured to sort the first candidate node subset from most to least according to the amount of remaining resources of the nodes, to obtain a second candidate node subset;

[0079] The computation module is used to build a bipartite graph based on the improved Hungarian algorithm, with the task description vector on the left and the second candidate node subset on the right.

[0080] Introducing slack variables for multi-resource constraints , establish the Lagrangian function:

[0081] ;

[0082] in, Representation node The remaining resources, Indicates the matching deviation between node resource efficiency and task description vector; Indicates the resources required by the task description vector;

[0083] A submodule is obtained, which is used to output the Top-5 candidate nodes through row and column reduction and augmented path search, and obtain the candidate node subset data.

[0084] 104. Fine-grained optimization module, used to perform fine-grained optimization on the candidate node subset data using the DRL-OPT dynamic reward and punishment reinforcement learning model, select the node that maximizes the long-term reward through Q-Learning, output the optimal node, and obtain resource allocation instructions;

[0085] Specifically, this embodiment also includes a model definition unit for defining the state space, action space and reward function of the DRL-OPT dynamic reward and punishment reinforcement learning model;

[0086] The node output unit is used to take the real-time state of the candidate node subset data as the input as the state space of the model, use the action space to generate a binary action of selection or non-selection for each node in the candidate node subset, and output a single optimal node;

[0087] The reward function unit is used to determine that the reward function includes a multi-dimensional dynamic reward and punishment mechanism, including at least positive rewards, negative rewards and dynamic adjustments.

[0088] State Space:

[0089] It takes the real-time status of a subset of candidate nodes as input, including:

[0090] Real-time data of the node resource matrix (normalized vectors such as CPU utilization, remaining memory, network bandwidth usage, and I / O latency);

[0091] Task requirement data (task description vector, urgency factor, data input and output scale);

[0092] Node historical scheduling records (recent task completion rate, average response time, number of resource allocation conflicts).

[0093] Action Space:

[0094] Generate a "select / not select" binary action for each node in the candidate node subset, and finally output a single optimal node (ensuring that only one node is assigned to execute a task each time).

[0095] Design a multi-dimensional dynamic reward and punishment mechanism to comprehensively evaluate long-term benefits: Positive rewards: task completion on time (based on urgency factor weighting), improved node resource utilization (avoiding excessive idleness), improved load balance (reduced load difference between adjacent nodes); Negative penalties: task timeout (inversely proportional to urgency), node overload (exceeding resource threshold), resource allocation conflict (repeated scheduling or insufficient resources); Dynamic adjustment: real-time correction of reward weights based on system load fluctuations (e.g., priority punishment of overload under high load, priority optimization of resource utilization under low load).

[0096] The action selection unit is used to select the node that maximizes the long-term reward through Q-Learning, collect the state data of the candidate nodes based on the candidate node subset data, and select the action through the ε-greedy strategy;

[0097] A greedy strategy unit, used for a greedy strategy that at least includes randomly exploring under-evaluated nodes with probability ε and selecting the node with the largest current Q value with probability 1-ε;

[0098] The reward calculation unit is used to calculate the immediate reward based on the actual results, update the Q value using the Q-Learning formula, output the optimal node, and obtain the resource allocation instruction;

[0099] ;

[0100] in, represents the state-action value function, represents the learning rate, represents the discount factor, Indicates the next state, Indicates the next action. Indicates that the status The maximum Q value that can be obtained from all possible actions.

[0101] 105. System scheduling management module, used to schedule the system according to resource allocation instructions, monitor the load of the optimal node, and start the resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

[0102] Specifically, this embodiment also includes:

[0103] The node management unit is used to encapsulate the optimal node ID and task parameters into a scheduling instruction and send it to the node manager. The node manager verifies the resource availability and starts the task execution if the verification is successful. Otherwise, an exception handling is triggered.

[0104] A threshold preset unit is used to preset multiple thresholds, including at least a warning threshold and an emergency threshold. The warning threshold is a CPU usage ≥ 80% and a memory usage ≥ 90% for 5 minutes; the emergency threshold is a CPU usage ≥ 95% and a memory usage ≥ 98% for 1 minute.

[0105] The resource borrowing and returning unit is used to start the resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

[0106] 1. Resource allocation instruction execution;

[0107] Scheduling execution:

[0108] The optimal node ID and task parameters (input data path, computing resource requirements) are encapsulated into a scheduling instruction and sent to the node manager;

[0109] The node manager verifies resource availability (based on the latest resource matrix). If the verification passes, the task execution is started. Otherwise, exception handling is triggered (return to step 3 for rematching).

[0110] 2. Load monitoring and threshold judgment;

[0111] Monitoring indicators:

[0112] Real-time collection of multi-dimensional load data of the optimal node:

[0113] Computing resources: CPU usage (average core load), memory usage (remaining physical memory / total amount);

[0114] Storage resources: disk I / O throughput, remaining storage space;

[0115] Network resources: inbound / outbound bandwidth usage, network latency (RTT).

[0116] Threshold setting:

[0117] Preset multi-level thresholds (warning threshold, emergency threshold), for example:

[0118] Warning threshold: CPU usage ≥ 80%, memory usage ≥ 90% for 5 minutes;

[0119] Emergency threshold: CPU usage ≥ 95%, memory usage ≥ 98% for 1 minute.

[0120] 3. Resource borrowing and returning protocol startup logic;

[0121] Trigger conditions:

[0122] When the real-time load of a node exceeds the warning threshold, resource pre-detection is triggered; when it exceeds the emergency threshold, the borrowing and returning process is immediately started.

[0123] Resource borrowing process:

[0124] Candidate node screening: Filter nodes from the system with a load lower than 50% and matching resource types (given priority to nodes in the same physical machine cluster to reduce network transmission overhead);

[0125] Borrowing Negotiation:

[0126] Send a resource borrowing request to the candidate node (including the type, quantity, and expected duration of the borrowed resource);

[0127] The candidate node responds whether it agrees based on its own resource scheduling strategy (such as reserving minimum resources);

[0128] Select the borrowing node based on the response priority (nodes with high resource surplus and high historical collaboration efficiency are given priority).

[0129] Resource transfer:

[0130] If computing resources are borrowed, partial load transfer can be achieved through container migration or task offloading.

[0131] If the borrowing is for storage / network resources, establish a temporary data channel or bandwidth allocation rule (to avoid affecting the original tasks of the borrowing node).

[0132] Resource return mechanism:

[0133] Set a borrowing timeout mechanism (e.g. the default borrowing time is no more than 30 minutes), and automatically trigger the return process when it expires;

[0134] When the optimal node load drops below the threshold (e.g., below 70% for 10 consecutive minutes), the borrowed resources are released proactively.

[0135] The borrower and the returner update the resource matrix and record the borrowing and return history (used for reward function calculation in step 4 to encourage collaboration between nodes).

[0136] 4. Fault tolerance and compensation in the borrowing and returning process;

[0137] If all candidate nodes refuse to borrow or a failure occurs during the transfer process, system-level resource reallocation is triggered:

[0138] Pause some processes of the current task to release overloaded node resources;

[0139] Return to step 3 to regenerate the candidate node subset, including more available nodes (including cross-cluster nodes).

[0140] Compensation rewards are given to nodes that provide resources (such as priority scheduling of their subsequent tasks and increase in resource quotas) to form a healthy collaborative ecosystem.

[0141] Its beneficial effects are: 1. By deploying monitoring agents at physical nodes to collect real-time electronic information data and constructing a resource matrix for normalization processing, it ensures that the basic data obtained by the system has high timeliness and standardized characteristics. The resource status information can truly reflect the node capabilities, provide a reliable basis for subsequent task matching, and avoid allocation deviations caused by data lag or format confusion. 2. Intelligent parsing and priority determination of task requirements parses the system task request to generate a task description vector, realizing the quantitative expression of task requirements. Dynamically distinguish between the urgency of tasks and the difference in resource requirements, give priority to guaranteeing the resource supply of high-priority tasks, and effectively improve the overall response efficiency and service quality of the system. 3. Efficient resource screening for coarse-grained matching Based on the improved Hungarian algorithm to traverse the initial information matrix data for coarse-grained matching, it can quickly screen out a subset of candidate nodes that meet the basic conditions of the task from a large number of nodes. Compared with traditional global search, this algorithm significantly reduces the computational complexity and shortens the task allocation decision time. 4. Fine-grained optimized dynamic resource adaptation utilizes the DRL-OPT dynamic reward and punishment reinforcement learning model combined with the Q-Learning algorithm to deeply optimize a subset of candidate nodes, selecting the optimal node with the goal of maximizing long-term rewards. It can adaptively learn the operating rules of the system and dynamically adjust the strategy based on historical allocation results, balancing resource utilization and task completion quality, avoiding local optimal traps, and achieving continuous evolution of resource allocation strategies. 5. Load balancing and elastic resource scheduling build a closed-loop dynamic resource management system by monitoring the load of the optimal node and initiating a resource borrowing and returning protocol. When the node load exceeds the threshold, the borrowing and returning protocol can automatically trigger resource migration or sharing, effectively preventing performance bottlenecks caused by single-point overload and ensuring system stability. At the same time, this mechanism supports elastic scaling of resources, reduces hardware redundancy costs, and improves resource economy.

[0142] In this embodiment, please refer to Figure 2 In a second embodiment of an electronic information dynamic scheduling and management system according to an embodiment of the present invention, the system data acquisition module includes the following submodules:

[0143] The data collection submodule is used to collect the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; computing resources include CPU utilization, GPU CUDA core occupancy and the number of idle FPGA computing units; storage resources include disk IOPS, remaining storage space and SSD life wear value; network resources include inter-node delay, remaining bandwidth and TCP retransmission rate; energy consumption data includes current power and energy consumption per unit computing power;

[0144] The matrix establishment submodule is used to establish a resource matrix in accordance with the node order based on the 4-dimensional indicators of each physical node in the real-time electronic information data;

[0145] The matrix normalization submodule is used to normalize the resources using the minimum-maximum normalization method, so that the indicators of each dimension are within the same value range and obtain the initial information matrix data.

[0146] The beneficial effect lies in deploying monitoring agents at physical nodes to collect real-time electronic information data and constructing a resource matrix for normalization. This ensures that the basic data obtained by the system is highly timely and standardized. This allows resource status information to truly reflect node capabilities, providing a reliable basis for subsequent task matching and avoiding allocation bias caused by data lag or formatting issues.

[0147] In this embodiment, please refer to Figure 3 In a third embodiment of an electronic information dynamic scheduling and management system according to an embodiment of the present invention, the fine-grained optimization module includes the following subunits:

[0148] The action selection unit is used to select the node that maximizes the long-term reward through Q-Learning, collect the state data of the candidate nodes based on the candidate node subset data, and select the action through the ε-greedy strategy;

[0149] A greedy strategy unit, used for a greedy strategy that at least includes randomly exploring under-evaluated nodes with probability ε and selecting the node with the largest current Q value with probability 1-ε;

[0150] The reward calculation unit is used to calculate the immediate reward based on the actual results, update the Q value using the Q-Learning formula, output the optimal node, and obtain the resource allocation instruction;

[0151] ;

[0152] in, represents the state-action value function, represents the learning rate, represents the discount factor, Indicates the next state, Indicates the next action. Indicates that the status The maximum Q value that can be obtained from all possible actions.

[0153] Its beneficial effect is to select the optimal node with the goal of maximizing long-term rewards. It can adaptively learn the operating rules of the system and dynamically adjust the strategy based on historical allocation results, balancing resource utilization and task completion quality, avoiding local optimal traps, and achieving continuous evolution of resource allocation strategies.

[0154] The above describes an electronic information dynamic scheduling and management system provided by an embodiment of the present invention. The following describes an electronic information dynamic scheduling and management method according to an embodiment of the present invention. Figure 4In one embodiment of the present invention, a method for dynamic scheduling and management of electronic information includes:

[0155] Step 401: Deploy a monitoring agent at each physical node to collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data;

[0156] Step 402: Receive a system task request, parse the system task request to generate a task description vector, and calculate the urgency factor to obtain task requirement data;

[0157] Step 403: Based on the improved Hungarian algorithm, the initial information matrix data is traversed and coarse-grained matching is performed according to the task description vector to obtain candidate node subset data;

[0158] Step 404: Use the DRL-OPT dynamic reward and punishment reinforcement learning model to perform fine-grained optimization on the candidate node subset data, select the node that maximizes the long-term reward through Q-Learning, output the optimal node, and obtain the resource allocation instruction;

[0159] Step 405: Schedule the system according to the resource allocation instruction, monitor the load of the optimal node, and start the resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

[0160] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An electronic information dynamic scheduling and management system, characterized in that: The electronic information dynamic scheduling and management system includes the following modules: A system data acquisition module is used to deploy a monitoring agent at each physical node, collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data; A task instruction receiving module is used to receive a system task request, parse the system task request to generate a task description vector, and calculate an urgency factor to obtain task requirement data; A coarse-grained matching module is used to traverse the initial information matrix data to perform coarse-grained matching based on the improved Hungarian algorithm according to the task description vector to obtain candidate node subset data; A fine-grained optimization module is used to perform fine-grained optimization on the candidate node subset data using the DRL-OPT dynamic reward and punishment reinforcement learning model, select the node that maximizes the long-term reward through Q-Learning, output the optimal node, and obtain resource allocation instructions; The system scheduling management module is used to schedule the system according to the resource allocation instruction, monitor the load of the optimal node, and start the resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

2. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The system data acquisition module includes the following submodules: The data collection submodule is used to collect the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; the computing resources include CPU utilization, GPU CUDA core occupancy rate and the number of idle FPGA computing units; the storage resources include disk IOPS, remaining storage space and SSD life wear value; the network resources include inter-node delay, remaining bandwidth and TCP retransmission rate; the energy consumption data includes current power and energy consumption per unit computing power; A matrix establishment submodule, configured to establish a resource matrix in node order based on the 4-dimensional index of each physical node in the real-time electronic information data; The matrix normalization submodule is used to normalize the resources using the minimum-maximum normalization method, so that the indicators of each dimension are within the same value range, and obtain initial information matrix data.

3. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The task instruction receiving module includes the following submodules: The task receiving submodule is used to receive a system task request, obtain the computational load, memory size, data volume, and deadline in the system task request, and obtain the task computation parameters; The urgency calculation submodule is used to integrate the task calculation parameters into a task description vector, obtain the current time and the deadline in the task description vector, and calculate the urgency factor of the task based on the pre-set weight coefficient and service quality level. ; ； in, and represents the pre-set weight coefficient, Indicates the current time, Indicates the deadline, Indicates the service quality level.

4. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The coarse-grained matching module includes the following submodules: A traversal submodule is used to traverse the initial information matrix data according to the computational memory requirements and data volume in the task description vector, screen out nodes that meet the basic requirements of the task for computational volume, memory, and data volume, and form the nodes into a first candidate node subset; a sorting submodule, configured to sort the first candidate node subset from most to least according to the amount of remaining resources of the nodes, to obtain a second candidate node subset; The computation module is used to build a bipartite graph based on the improved Hungarian algorithm, with the task description vector on the left and the second candidate node subset on the right. Introducing slack variables for multi-resource constraints , establish the Lagrangian function: ； in, Representation node The remaining resources, Indicates the matching deviation between node resource efficiency and task description vector; Indicates the resources required by the task description vector; A submodule is obtained, which is used to output the Top-5 candidate nodes through row and column reduction and augmented path search, and obtain the candidate node subset data.

5. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The fine-grained optimization module includes the following units: Model definition unit, used to define the state space, action space and reward function of the DRL-OPT dynamic reward and punishment reinforcement learning model; a node output unit, configured to take the real-time state of the candidate node subset data as input as the state space of the model, generate a binary action of selection or non-selection for each node in the candidate node subset using the action space, and output a single optimal node; The reward function unit is used to determine that the reward function includes a multi-dimensional dynamic reward and punishment mechanism, including at least positive rewards, negative rewards and dynamic adjustments.

6. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The fine-grained optimization module also includes the following units: An action selection unit, configured to select a node that maximizes a long-term reward through Q-Learning, collect state data of candidate nodes based on the candidate node subset data, and select an action through an ε-greedy strategy; A greedy strategy unit, configured to implement a greedy strategy that at least includes randomly exploring under-evaluated nodes with a probability of ε and selecting a node with the largest current Q value with a probability of 1-ε; The reward calculation unit is used to calculate the immediate reward based on the actual results, update the Q value using the Q-Learning formula, output the optimal node, and obtain the resource allocation instruction; ； in, represents the state-action value function, represents the learning rate, represents the discount factor, Indicates the next state, Indicates the next action. Indicates that the status The maximum Q value that can be obtained from all possible actions.

7. The electronic information dynamic scheduling and management system according to claim 1, characterized in that: The system scheduling management module includes the following units: The node management unit is used to encapsulate the optimal node ID and task parameters into a scheduling instruction and send it to the node manager. The node manager verifies the resource availability and starts the task execution if the verification is successful. Otherwise, an exception handling is triggered. A threshold preset unit is used to preset multiple thresholds, including at least a warning threshold and an emergency threshold. The warning threshold is ≥80% CPU usage and ≥90% memory usage for 5 minutes. The emergency threshold is CPU usage ≥ 95%, memory usage ≥ 98%, for 1 minute; The resource borrowing and returning unit is configured to start a resource borrowing and returning protocol when the real-time load of the optimal node exceeds a preset threshold.

8. A method for dynamic scheduling and management of electronic information, characterized in that: The electronic information dynamic scheduling and management method comprises the following steps: Deploy a monitoring agent at each physical node to collect real-time electronic information data, construct a resource matrix based on the real-time electronic information data, and perform normalization processing to obtain initial information matrix data; Receiving a system task request, parsing the system task request to generate a task description vector, and calculating an urgency factor to obtain task requirement data; Based on the improved Hungarian algorithm, the initial information matrix data is traversed to perform coarse-grained matching according to the task description vector to obtain candidate node subset data; The DRL-OPT dynamic reward and punishment reinforcement learning model is used to perform fine-grained optimization on the candidate node subset data, and the node that maximizes the long-term reward is selected through Q-Learning. The optimal node is output and the resource allocation instruction is obtained; The system is scheduled according to the resource allocation instruction, and the load of the optimal node is monitored. When the real-time load of the optimal node exceeds a preset threshold, the resource borrowing and returning protocol is started.

9. A method for dynamic scheduling and management of electronic information as claimed in claim 8, characterized in that: The monitoring agent is deployed at each physical node to collect real-time electronic information data, a resource matrix is ​​constructed based on the real-time electronic information data, and normalized to obtain initial information matrix data, including: Collect the usage of computing resources, storage resources, network resources and energy consumption data of physical nodes to obtain real-time electronic information data; the computing resources include CPU utilization, GPU CUDA core occupancy rate and the number of idle FPGA computing units; the storage resources include disk IOPS, remaining storage space and SSD life wear value; the network resources include inter-node delay, bandwidth remaining amount and TCP retransmission rate; the energy consumption data includes current power and energy consumption per unit computing power; Establishing a resource matrix according to the order of nodes based on the 4-dimensional index of each physical node in the real-time electronic information data; The resources are normalized using the minimum-maximum normalization method so that the indicators of each dimension are within the same value range to obtain the initial information matrix data.

10. The method for dynamic scheduling and management of electronic information according to claim 8, characterized in that: The receiving of the system task request, parsing the system task request to generate a task description vector, and calculating the urgency factor to obtain task requirement data includes: Receive a system task request, obtain the computational load, memory size, data volume, and deadline in the system task request, and obtain task computation parameters; Integrate the task calculation parameters into a task description vector, obtain the current time and the deadline in the task description vector, and calculate the urgency factor of the task based on the pre-set weight coefficient and the service quality level ; ； in, and represents the pre-set weight coefficient, Indicates the current time, Indicates the deadline, Indicates the service quality level.

Images