Calculation power scheduling method and device

Abstract

The invention relates to the field of artificial intelligence, and discloses a computing power scheduling method and device, and the method comprises the following steps: S01, building a computing power network resource library, and storing computing power resources in a unified manner, S02, marking computing power network nodes, and collecting computing power resource management information, and S03, building a priority comprehensive model to calculate the priority index of each task, s04, determining an execution sequence of the tasks according to the priority indexes of the tasks, S05, carrying out resource load calculation on the nodes to obtain resource load indexes of the nodes, and S06, carrying out resource load judgment on the nodes according to the resource load indexes of the nodes, s07, task execution operation is carried out on the to-be-processed task waiting sequence according to the resource load judgment result, S08, computing power scheduling of all nodes is completed, according to the computing power scheduling method and device, computing power resources are reasonably distributed, it is ensured that key tasks are processed preferentially, and the data processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and more specifically to a computing power scheduling method and device. Background Art

[0002] In the world or in a specific region, the distribution of computing resources is often uneven. Some regions or data centers may have a large amount of computing resources, while other regions may be relatively scarce. This imbalance leads to waste and inefficiency of computing resources. Therefore, an effective computing power scheduling method is needed to optimize the allocation and utilization of resources. There are significant differences in the demand for computing power in different industries and application scenarios. For example, fields such as artificial intelligence, big data analysis, and high-performance computing have high demands for computing power, while traditional applications have relatively low demands. This difference in demand has also prompted the emergence of computing power scheduling methods to meet the computing power needs in different scenarios.

[0003] However, existing computing power scheduling methods lack effective management and monitoring of computing power resources, resulting in problems such as insufficient resource utilization, resource waste or resource overload. The algorithm lacks sufficient flexibility and intelligence in resource allocation and task scheduling, and cannot fully adapt to dynamically changing computing power requirements and resource conditions. The lack of priority processing causes computing power resources to be occupied by low-priority tasks for a long time, while high-priority tasks cannot be processed in a timely manner, which not only reduces resource utilization, but also affects the overall system performance and user experience. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a computing power scheduling method and device to solve the problems existing in the above-mentioned background technology.

[0005] The present invention provides the following technical solution: a computing power scheduling method, comprising the following steps:

[0006] Step S01: Establish a computing power network resource library to store computing power resources in a unified manner: through resource pooling technology, various computing power network resources are stored in the resource library in a virtualized form to form a unified computing power resource pool;

[0007] Step S02: Marking computing network nodes and collecting computing resource management information: obtaining node identification, adding node identification and type information to a predefined link state announcement, and marking the link state announcement of the computing network node through a computing network configured with a preset routing protocol;

[0008] Step S03: Establishing a priority comprehensive model to calculate the priority index of each task: Based on the task information to be processed collected in step S02, establishing a priority comprehensive model to calculate the priority index of each task;

[0009] Step S04: determining the execution order of tasks according to the priority index of each task: determining the execution order of tasks based on the priority index of each task calculated in step S03, forming a waiting sequence of tasks to be processed;

[0010] Step S05: Calculate the resource load of each node to obtain the resource load index of each node: Calculate the resource load index of each node based on the resource load information of each node collected in step S02;

[0011] Step S06: judging the resource load of each node according to the resource load index of each node: comparing the resource load index of each node with a preset node resource load threshold, and judging the resource load of each node;

[0012] Step S07: performing task execution operation on the waiting sequence of tasks to be processed according to the result of resource load judgment: marking executable task nodes according to the result of resource load judgment in step S06, and processing the waiting sequence of tasks to be processed formed in step S04;

[0013] Step S08: Complete the computing power scheduling of each node: store the scheduled computing power resource management information in the computing power network resource library.

[0014] Preferably, in step S02, the computing power network nodes are marked and computing power resource management information is collected. The computing power resource management information includes information on pending tasks and resource load information of each node. The pending task information includes the number of network tasks and factors affecting task allocation. The resource load information of each node includes the length of the task queue currently waiting to be processed at the node, the total amount of current tasks at the node, the resources currently used by the node, and the total resources of the node.

[0015] Preferably, in step S03, based on the task information to be processed collected in step S02, a priority comprehensive model is established to calculate the priority index of each task. The specific contents are as follows:

[0016] Step S11: selecting the entire CPU cycle required for the task, the size of the task data, the safety requirement level of the task, and the maximum delay constraint of the task as four influencing factors of the task;

[0017] Step S12: construct a comparison matrix A, expressed as: A = (a pq ) n×n , a pq represents the elements in the matrix, and n represents the number of influencing factors of the selected task;

[0018] Step S13: The elements in the matrix are expressed as

[0019] Step S14: Assign a weight to each task under each factor, and the calculation formula of the vector corresponding to the weight is: Where r represents the network task, k represents the number of factors affecting task allocation;

[0020] Step S15: construct the vector corresponding to the task weight into a matrix, expressed as: Where B represents the vector matrix corresponding to the task weight;

[0021] Step S16: The eigenvalue vector of the vector matrix corresponding to the task weight is expressed as: D = [λ1, λ2, λ3, λ4] T ,in Where k = 1, 2, 3, 4, λ represents the eigenvalue vector of the vector matrix corresponding to the task weight;

[0022] Step S17: Calculate the priority index of each task, the calculation formula is: D = B × λ, where D represents the priority index of each task, B represents the vector matrix corresponding to the task weight, and λ represents the eigenvalue vector of the vector matrix corresponding to the task weight.

[0023] Preferably, in step S04, the execution order of tasks is determined based on the priority index of each task calculated in step S03, and the specific content of forming a waiting sequence of tasks to be processed is: all tasks are sorted from high to low according to the priority index; if there are multiple tasks with the same priority index, they are subdivided and sorted according to the task reception time, and an estimated execution time is assigned to each task to form a waiting sequence of tasks to be processed.

[0024] Preferably, in step S05, the specific content of calculating the resource load index of each node based on the resource load information of each node collected in step S02 is as follows:

[0025] Step S11: Calculate the waiting task processing rate of each node, and the calculation formula is: Where y1 represents the waiting task processing rate of each node, L1 represents the length of the task queue currently waiting to be processed at the node, and L represents the total number of tasks currently at the node;

[0026] Step S12: Calculate the resource utilization of each node, the calculation formula is: Where y2 represents the resource utilization of each node, R2 represents the resources currently used by the node, and R represents the total resources of the node;

[0027] Step S13: Calculate the resource load index of each node, the calculation formula is: α=ln(c1×y1+c2×y2), where α represents the resource load index of each node, c1 and c2 represent constants, y1 represents the waiting task processing rate of each node, and y2 represents the resource utilization rate of each node;

[0028] Step S14: Compare the resource load index of each node with the preset node load threshold. If the resource load index is greater than or equal to the preset node load threshold, the node is judged to be overloaded and task allocation management is not prioritized. If the resource load index is less than the preset node load threshold, the node is judged to be normal and task allocation management is prioritized.

[0029] Preferably, in step S07, the executable task nodes are marked according to the resource load judgment result in step S06, and the specific content of processing the waiting sequence of pending tasks formed in step S04 is: marking the nodes for priority task allocation management according to the resource load judgment result, and allocating the waiting sequence of pending tasks formed in step S04 to the marked nodes in priority.

[0030] Preferably, in step S08, the computing nodes to which each task is assigned, the allocation time, and the required amount of resources are recorded in detail, the computing resources are dynamically scheduled according to the priority of the task, and the scheduled computing resource management information is stored in the computing network resource library.

[0031] A computing power scheduling device, comprising an electronic device and a computer-readable storage medium;

[0032] The electronic device comprises a processor, a memory and a communication bus, wherein the memory stores a computer-readable program executable by the processor;

[0033] The communication bus realizes the connection and communication between the processor and the memory;

[0034] When the processor executes the computer-readable program, steps in a computing power scheduling method are implemented;

[0035] The computer-readable storage medium stores one or more programs that can be executed by one or more processors.

[0036] Technical effects and advantages of the present invention:

[0037] The present invention is provided with step S01: establishing a computing power network resource library to uniformly store computing power resources, step S02: marking computing power network nodes and collecting computing power resource management information, step S03: establishing a priority comprehensive model to calculate the priority index of each task, step S04: determining the execution order of tasks according to the priority index of each task, step S05: calculating the resource load of each node to obtain the resource load index of each node, step S06: judging the resource load of each node according to the resource load index of each node, step S07: performing task execution operations on the waiting sequence of tasks to be processed according to the resource load judgment results, and step S08: completing the computing power scheduling of each node. A computing power scheduling method and device ensure priority processing of key tasks and improve data processing efficiency by reasonably allocating computing power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The figure is a flow chart of a computing power scheduling method. DETAILED DESCRIPTION

[0039] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely illustrative. The computing power scheduling method involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0040] like Figure 1 As shown, the present invention provides a computing power scheduling method, comprising the following steps:

[0041] Step S01: Establish a computing power network resource library to store computing power resources in a unified manner: through resource pooling technology, various computing power network resources are stored in the resource library in a virtualized form to form a unified computing power resource pool;

[0042] Step S02: Marking computing network nodes and collecting computing resource management information: obtaining node identification, adding node identification and type information to a predefined link state announcement, and marking the link state announcement of the computing network node through a computing network configured with a preset routing protocol;

[0043] Step S03: Establishing a priority comprehensive model to calculate the priority index of each task: Based on the task information to be processed collected in step S02, establishing a priority comprehensive model to calculate the priority index of each task;

[0044] Step S04: determining the execution order of tasks according to the priority index of each task: determining the execution order of tasks based on the priority index of each task calculated in step S03, forming a waiting sequence of tasks to be processed;

[0045] Step S05: Calculate the resource load of each node to obtain the resource load index of each node: Calculate the resource load index of each node based on the resource load information of each node collected in step S02;

[0046] Step S06: judging the resource load of each node according to the resource load index of each node: comparing the resource load index of each node with a preset node resource load threshold, and judging the resource load of each node;

[0047] Step S07: performing task execution operation on the waiting sequence of tasks to be processed according to the result of resource load judgment: marking executable task nodes according to the result of resource load judgment in step S06, and processing the waiting sequence of tasks to be processed formed in step S04;

[0048] Step S08: Complete the computing power scheduling of each node: store the scheduled computing power resource management information in the computing power network resource library.

[0049] In this embodiment, it should be specifically explained that in step S02, the computing power network nodes are marked and computing power resource management information is collected. The computing power resource management information includes information on pending tasks and resource load information of each node. The pending task information includes the number of network tasks and factors affecting task allocation. The resource load information of each node includes the length of the task queue currently waiting to be processed at the node, the total number of current tasks at the node, the resources currently used by the node, and the total resources of the node.

[0050] In this embodiment, it should be specifically explained that in step S03, based on the task information to be processed collected in step S02, the specific contents of establishing a priority comprehensive model to calculate the priority index of each task are as follows:

[0051] Step S11: selecting the entire CPU cycle required for the task, the size of the task data, the safety requirement level of the task, and the maximum delay constraint of the task as four influencing factors of the task;

[0052] Step S12: construct a comparison matrix A, expressed as: A = (a pq ) n×n , a pq represents the elements in the matrix, and n represents the number of influencing factors of the selected task;

[0053] Step S13: The elements in the matrix are expressed as

[0054] Step S14: Assign a weight to each task under each factor, and the calculation formula of the vector corresponding to the weight is: Where r represents the network task, k represents the number of factors affecting task allocation;

[0055] Step S15: construct the vector corresponding to the task weight into a matrix, expressed as: Where B represents the vector matrix corresponding to the task weight;

[0056] Step S16: The eigenvalue vector of the vector matrix corresponding to the task weight is expressed as: D = [λ1, λ2, λ3, λ4] T ,in Where k = 1, 2, 3, 4, λ represents the eigenvalue vector of the vector matrix corresponding to the task weight;

[0057] Step S17: Calculate the priority index of each task, the calculation formula is: D = B × λ, where D represents the priority index of each task, B represents the vector matrix corresponding to the task weight, and λ represents the eigenvalue vector of the vector matrix corresponding to the task weight.

[0058] In this embodiment, it should be specifically explained that in step S04, the execution order of tasks is determined based on the priority index of each task calculated in step S03, and the specific content of forming the waiting sequence of tasks to be processed is: all tasks are sorted from high to low according to the priority index. If there are multiple tasks with the same priority index, they are subdivided and sorted according to the task reception time, and an estimated execution time is assigned to each task to form a waiting sequence of tasks to be processed.

[0059] In this embodiment, it should be specifically explained that in step S05, the specific content of the resource load index of each node calculated based on the resource load information of each node collected in step S02 is as follows:

[0060] Step S11: Calculate the waiting task processing rate of each node, and the calculation formula is: Where y1 represents the waiting task processing rate of each node, L1 represents the length of the task queue currently waiting to be processed at the node, and L represents the total number of tasks currently at the node;

[0061] Step S12: Calculate the resource utilization of each node, the calculation formula is: Where y2 represents the resource utilization of each node, R2 represents the resources currently used by the node, and R represents the total resources of the node;

[0062] Step S13: Calculate the resource load index of each node, the calculation formula is: α=ln(c1×y1+c2×y2), where α represents the resource load index of each node, c1 and c2 represent constants, y1 represents the waiting task processing rate of each node, and y2 represents the resource utilization rate of each node;

[0063] Step S14: Compare the resource load index of each node with the preset node load threshold. If the resource load index is greater than or equal to the preset node load threshold, the node is judged to be overloaded and task allocation management is not prioritized. If the resource load index is less than the preset node load threshold, the node is judged to be normal and task allocation management is prioritized.

[0064] In this embodiment, it should be specifically explained that in step S07, the executable task nodes are marked according to the resource load judgment result in step S06, and the specific content of processing the waiting sequence of pending tasks formed in step S04 is: marking the nodes for priority task allocation management according to the resource load judgment result, and allocating the waiting sequence of pending tasks formed in step S04 to the marked nodes in priority.

[0065] In this embodiment, it should be specifically explained that in step S08, the computing node to which each task is assigned, the allocation time and the required amount of resources are recorded in detail, the computing resources are dynamically scheduled according to the priority of the task, and the scheduled computing resource management information is stored in the computing network resource library.

[0066] A computing power scheduling device, comprising an electronic device and a computer-readable storage medium;

[0067] The electronic device comprises a processor, a memory and a communication bus, wherein the memory stores a computer-readable program executable by the processor;

[0068] The communication bus realizes the connection and communication between the processor and the memory;

[0069] When the processor executes the computer-readable program, steps in a computing power scheduling method are implemented;

[0070] The computer-readable storage medium stores one or more programs that can be executed by one or more processors.

[0071] The difference between this embodiment and the prior art mainly lies in that this embodiment is provided with step S01: establishing a computing power network resource library to uniformly store computing power resources, step S02: marking computing power network nodes and collecting computing power resource management information, step S03: establishing a priority comprehensive model to calculate the priority index of each task, step S04: determining the execution order of tasks according to the priority index of each task, step S05: calculating the resource load of each node to obtain the resource load index of each node, step S06: judging the resource load of each node according to the resource load index of each node, step S07: performing task execution operations on the waiting sequence of tasks to be processed according to the result of resource load judgment, step S08: completing computing power scheduling of each node, a computing power scheduling method and device ensure priority processing of key tasks and improve data processing efficiency by reasonably allocating computing power resources.

[0072] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0073] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A computing power scheduling method, characterized in that: The following steps are involved: Step S01: Establish a computing power network resource library to store computing power resources in a unified manner: through resource pooling technology, various computing power network resources are stored in the resource library in a virtualized form to form a unified computing power resource pool; Step S02: Marking computing network nodes and collecting computing resource management information: obtaining node identification, adding node identification and type information to a predefined link state announcement, and marking the link state announcement of the computing network node through a computing network configured with a preset routing protocol; Step S03: Establishing a priority comprehensive model to calculate the priority index of each task: Based on the task information to be processed collected in step S02, establishing a priority comprehensive model to calculate the priority index of each task; Step S04: determining the execution order of tasks according to the priority index of each task: determining the execution order of tasks based on the priority index of each task calculated in step S03, forming a waiting sequence of tasks to be processed; Step S05: Calculate the resource load of each node to obtain the resource load index of each node: Calculate the resource load index of each node based on the resource load information of each node collected in step S02; Step S06: judging the resource load of each node according to the resource load index of each node: comparing the resource load index of each node with a preset node resource load threshold, and judging the resource load of each node; Step S07: performing task execution operation on the waiting sequence of tasks to be processed according to the result of resource load judgment: marking executable task nodes according to the result of resource load judgment in step S06, and processing the waiting sequence of tasks to be processed formed in step S04; Step S08: Complete the computing power scheduling of each node: store the scheduled computing power resource management information in the computing power network resource library.

2. A computing power scheduling method according to claim 1, characterized in that: In the step S02, the computing power network nodes are marked and computing power resource management information is collected. The computing power resource management information includes information on pending tasks and resource load information of each node. The pending task information includes the number of network tasks and factors affecting task allocation. The resource load information of each node includes the length of the task queue currently waiting to be processed at the node, the total amount of current tasks at the node, the resources currently used by the node, and the total resources of the node.

3. A computing power scheduling method according to claim 1, characterized in that: In step S03, based on the task information to be processed collected in step S02, a priority comprehensive model is established to calculate the priority index of each task. The specific contents are as follows: Step S11: selecting the entire CPU cycle required for the task, the size of the task data, the safety requirement level of the task, and the maximum delay constraint of the task as four influencing factors of the task; Step S12: construct a comparison matrix A, expressed as: A = (a pq ) n′n , a pq represents the elements in the matrix, and n represents the number of influencing factors of the selected task; Step S13: The elements in the matrix are expressed as Step S14: Assign a weight to each task under each factor, and the calculation formula of the vector corresponding to the weight is: Where r represents the network task, k represents the number of factors affecting task allocation; Step S15: construct the vector corresponding to the task weight into a matrix, expressed as: Where B represents the vector matrix corresponding to the task weight; Step S16: The eigenvalue vector of the vector matrix corresponding to the task weight is expressed as: D = [λ1, λ2, λ3, λ4] T ,in Where k = 1, 2, 3, 4, λ represents the eigenvalue vector of the vector matrix corresponding to the task weight; Step S17: Calculate the priority index of each task, the calculation formula is: D = B × λ, where D represents the priority index of each task, B represents the vector matrix corresponding to the task weight, and λ represents the eigenvalue vector of the vector matrix corresponding to the task weight.

4. A computing power scheduling method according to claim 1, characterized in that: In the step S04, the execution order of the tasks is determined based on the priority index of each task calculated in step S03, and the specific content of forming the waiting sequence of tasks to be processed is: all tasks are sorted from high to low according to the priority index. If there are multiple tasks with the same priority index, they are subdivided and sorted according to the task reception time, and an estimated execution time is assigned to each task to form a waiting sequence of tasks to be processed.

5. A computing power scheduling method according to claim 1, characterized in that: In step S05, the specific content of the resource load index of each node calculated based on the resource load information of each node collected in step S02 is as follows: Step S11: Calculate the waiting task processing rate of each node, and the calculation formula is: Where y1 represents the waiting task processing rate of each node, L1 represents the length of the task queue currently waiting to be processed at the node, and L represents the total number of tasks currently at the node; Step S12: Calculate the resource utilization of each node, the calculation formula is: Where y2 represents the resource utilization of each node, R2 represents the resources currently used by the node, and R represents the total resources of the node; Step S13: Calculate the resource load index of each node, the calculation formula is: α=ln(c1′y1+c2′y2), where α represents the resource load index of each node, c1 and c2 represent constants, y1 represents the waiting task processing rate of each node, and y2 represents the resource utilization rate of each node; Step S14: Compare the resource load index of each node with the preset node load threshold. If the resource load index is greater than or equal to the preset node load threshold, the node is judged to be overloaded and task allocation management is not prioritized. If the resource load index is less than the preset node load threshold, the node is judged to be normal and task allocation management is prioritized.

6. A computing power scheduling method according to claim 1, characterized in that: In the step S07, the executable task nodes are marked according to the resource load judgment result in step S06, and the specific content of processing the waiting sequence of pending tasks formed in step S04 is: marking the nodes for priority task allocation management according to the resource load judgment result, and allocating the waiting sequence of pending tasks formed in step S04 to the marked nodes in priority.

7. A computing power scheduling method according to claim 1, characterized in that: In step S08, the computing nodes to which each task is assigned, the allocation time, and the required amount of resources are recorded in detail, the computing resources are dynamically scheduled according to the priority of the task, and the scheduled computing resource management information is stored in the computing network resource library.

8. A computing power scheduling device, characterized in that: It is implemented based on a computing power scheduling method according to any one of claims 1 to 7, and includes an electronic device and a computer-readable storage medium; The electronic device comprises a processor, a memory and a communication bus, wherein the memory stores a computer-readable program executable by the processor; The communication bus realizes the connection and communication between the processor and the memory; When the processor executes the computer-readable program, the steps in the computing power scheduling method according to any one of claims 1 to 7 are implemented; The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in a computing power scheduling method as described in any one of claims 1-7.

Images