Multi-domain distributed scheduling method and system of scheduler

Abstract

The invention provides a multi-domain distributed scheduling method and system of a scheduler. The multi-domain distributed scheduling method comprises the following steps: acquiring a job with the highest priority from a to-be-scheduled task list; acquiring queue information of the job with the highest priority; wherein each queue in the queue information is configured with a plurality of plug-ins and distributed with a plurality of nodes; monitoring information of each node is acquired; wherein the monitoring information comprises a CPU utilization rate, a memory utilization rate, a disk utilization rate and a network utilization rate; calculating scoring results corresponding to all plug-ins according to the monitoring information of each node; the scoring results corresponding to all the plug-ins are analyzed through a multi-index decision analysis algorithm, and a comprehensive score is obtained; selecting a node with the highest score from the comprehensive scores as a scheduling node of the to-be-scheduled task, and executing job scheduling; the cluster resource utilization rate on job scheduling can be improved, load balancing is achieved, and the scheduling requirement in a specific scene is met.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a multi-domain distributed scheduling method and system for a scheduler. Background Art

[0002] On a computing power cluster with a distributed architecture, different types of computing tasks and services in fields such as big data, scientific computing, machine learning, and large models need to be run.

[0003] The Kubernetes scheduler runs on the Master node and is a pluggable module. Its scheduling algorithm is relatively simple, the cluster resource utilization rate is not high, the load balancing is poor, and it cannot meet the scheduling requirements in specific scenarios. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a multi-domain distributed scheduling method and system for a scheduler, which can improve the cluster resource utilization rate in job scheduling, achieve load balancing, and meet the scheduling requirements in specific scenarios.

[0005] In a first aspect, an embodiment of the present invention provides a multi-domain distributed scheduling method for a scheduler, and the method includes: Obtain the job with the highest priority from the list of tasks to be scheduled; Obtain the queue information of the job with the highest priority; wherein, multiple plugins are configured for each queue in the queue information and multiple nodes are allocated; Obtain the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate, and network utilization rate; Calculate the scoring results corresponding to all the plugins according to the monitoring information of each node; Analyze the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; Select the node with the highest score from the comprehensive score as the scheduling node for the task to be scheduled, and perform job scheduling.

[0006] Further, calculating the scoring results corresponding to all the plugins according to the monitoring information of each node includes repeatedly performing the following processing until each plugin is traversed: Load the first plugin, and call the screening strategy interface of the first plugin to filter out the nodes that do not meet the requirements; Call the scoring strategy interface of the first plugin to score the nodes that meet the requirements to obtain a first scoring result.

[0007] Further, calling the scoring strategy interface of the first plugin to score the nodes that meet the requirements to obtain a first scoring result includes: Calculate the first scoring result according to the following formula:

[0008] where WorstFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resource of the i-th node, free_mem is the remaining memory resource of the i-th node, and k1 and k2 are constant coefficients.

[0009] Further, call the scoring strategy interface of the first plugin to score the nodes that meet the requirements, and obtain the first scoring result, including: Calculate the first scoring result according to the following formula:

[0010] where BestFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resource of the i-th node, free_mem is the remaining memory resource of the i-th node, free_disk is the remaining disk resource of the i-th node, free_net is the remaining network resource of the i-th node, request_cpu is the CPU resource requested for task running of the i-th node, request_mem is the memory resource requested for task running of the i-th node, request_disk is the disk resource requested for task running of the i-th node, request_net is the network resource requested for task running of the i-th node, and w1, w2, w3, and w4 are constant coefficients.

[0011] Further, analyze the scoring results corresponding to all the plugins through a multi-criteria decision-making analysis algorithm to obtain a comprehensive score, including: Construct a score matrix from the scoring results corresponding to all the plugins; Obtain the comprehensive score by using a weighted summation algorithm for the score matrix.

[0012] Further, obtain the job with the highest priority from the list of tasks to be scheduled, including: Obtain the list of tasks to be scheduled; Traverse the jobs in the list of tasks to be scheduled according to the task priority order to obtain the job with the highest priority.

[0013] Further, the method further includes: If all the nodes do not meet the scheduling conditions, the tasks to be scheduled are not scheduled.

[0014] Second aspect, an embodiment of the present invention provides a multi-domain distributed scheduling system for a scheduler. The system includes: A job acquisition module, configured to acquire the job with the highest priority from the list of tasks to be scheduled; A queue information acquisition module, configured to acquire the queue information of the job with the highest priority; wherein, multiple plugins are configured for each queue in the queue information and multiple nodes are allocated; A monitoring information acquisition module, configured to acquire the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate, and network utilization rate; A calculation module, configured to calculate the scoring results corresponding to all the plugins according to the monitoring information of each node; An analysis module, configured to analyze the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; A selection module, configured to select the node with the highest score from the comprehensive score as the scheduling node of the task to be scheduled, and perform job scheduling.

[0015] Third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor. A computer program that can run on the processor is stored on the memory. When the processor executes the computer program, the method described above is implemented.

[0016] Fourth aspect, an embodiment of the present invention provides a computer-readable medium having non-volatile program code executable by a processor. The program code causes the processor to execute the method described above.

[0017] An embodiment of the present invention provides a multi-domain distributed scheduling method and system for a scheduler, including: acquiring the job with the highest priority from the list of tasks to be scheduled; acquiring the queue information of the job with the highest priority; wherein, multiple plugins are configured for each queue in the queue information and multiple nodes are allocated; acquiring the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate, and network utilization rate; calculating the scoring results corresponding to all the plugins according to the monitoring information of each node; analyzing the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; selecting the node with the highest score from the comprehensive score as the scheduling node of the task to be scheduled, and performing job scheduling; which can improve the utilization rate of cluster resources in job scheduling, achieve load balancing, and meet the scheduling requirements in specific scenarios.

[0018] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the specification, the claims, and the drawings.

[0019] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following provides preferred embodiments in conjunction with the accompanying drawings and describes them in detail as follows. Brief Description of the Drawings

[0020] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a flowchart of the multi-domain distributed scheduling method of the scheduler provided in the first embodiment of the present invention; Figure 2 It is a signaling diagram of the scheduler working provided in the first embodiment of the present invention; Figure 3 It is a schematic diagram of the architecture of the scheduler provided in the first embodiment of the present invention; Figure 4 It is a schematic diagram of the queue design provided in the first embodiment of the present invention; Figure 5 It is a schematic diagram of the multi-domain distributed scheduling system of the scheduler provided in the second embodiment of the present invention. Specific Embodiments

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0023] The scheduling requirements of the distributed architecture scheduler include: job scheduling, service scheduling, and resource scheduling.

[0024] The running time of job scheduling is usually from a few minutes to several hours. Job scheduling mainly focuses on: 1) throughput, the number of tasks that can be processed per unit time; 2) execution time, the execution time from start to end; 3) the length of the waiting queue, the number of tasks that have arrived but have not been scheduled.

[0025] Service scheduling means that once a service starts running, it will continue to run for a relatively long time (several months). Service scheduling mainly focuses on: 1) Timing control, without affecting the correctness of the running service; 2) Concurrency control, without affecting the SLA of the running service; 3) Service availability, try to disperse to different error domains (regions, computer rooms, switches) to reduce the impact range when a failure occurs; 4) Load balancing; 5) As the service changes, the distribution of cluster resources will show a certain degree of imbalance and needs to be coordinated; 6) Rebalance mechanism to reduce resource fragmentation and improve cluster capacity.

[0026] As the common part of job scheduling and service scheduling, the core issues that resource scheduling focuses on are: 1) Resource utilization rate: when the input remains unchanged, reduce the number of machines used; 2) Resource fragmentation: when the total resources remain unchanged, improve the loading capacity of the cluster. This application schedules tasks in multiple computing fields (big data, scientific computing, machine learning, large models) at the same time to solve the problem of cluster utilization rate in job scheduling and how to balance service quality.

[0027] To facilitate the understanding of this embodiment, the embodiments of the present invention will be introduced in detail below.

[0028] Embodiment 1: Figure 1 It is a flowchart of the multi-domain distributed scheduling method of the scheduler provided in Embodiment 1 of the present invention.

[0029] Refer to Figure 1 and the method includes the following steps: Step S101, obtain the job with the highest priority from the list of tasks to be scheduled; Step S102, obtain the queue information of the job with the highest priority; wherein, each queue in the queue information configures multiple plugins and allocates multiple nodes; Step S103, obtain the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate and network utilization rate; Specifically, each queue is pre-configured with a series of plugins (scheduling policies) by the cluster administrator to obtain all node information, and this information is collectively referred to as the scheduling context.

[0030] Step S104, calculate the scoring results corresponding to all plugins according to the monitoring information of each node; Step S105, analyze the scoring results corresponding to all plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; Step S106, select the node with the highest score from the comprehensive score as the scheduling node for the task to be scheduled; Step S107, execute job scheduling.

[0031] Specifically, traverse all plugins. If there are plugins, load the plugins and call the filtering policy interface and scoring policy interface of the plugins. If there are no plugins, execute step S106. After performing job scheduling, continue to repeat step S101.

[0032] Refer to Figure 2 , part of the Scheduler.

[0033] API Server: The central management hub of Kubernetes, the entry for all cluster operations.

[0034] Etcd: A distributed key-value storage database that stores all state data of the Kubernetes cluster.

[0035] Scheduler: Determines which node a Pod should run on. The Scheduler reads / updates data through the API Server and does not directly operate on etcd or OCI.

[0036] kubelete: The "agent" on the node, responsible for performing the lifecycle management of Pods (such as creating / destroying containers).

[0037] OCI: Defines the standard specifications for container runtimes and images (such as runc, containerd). The OCI standard is called by kubelet (instead of Scheduler / API Server), and containers are actually run on the node.

[0038] Refer to Figure 3 , the architecture of the scheduler includes a scheduling service and a series of pluggable plugins, and the scheduling data is provided by the cluster monitoring service.

[0039] Refer to Figure 4 , a queue can include N nodes. The same node can be assigned to different queues at the same time. Also, to handle services with daily access tidal characteristics, the queue is designed as a tidal queue, that is, the nodes in the queue can change with the time period during the day. The queue contains a series of plugins (scheduling policies) to handle different scheduling requirements.

[0040] Furthermore, step S104 includes the following steps. Repeat the following processing until each plugin has been traversed: Step S201, load the first plugin and call the filtering policy interface of the first plugin to filter out nodes that do not meet the requirements; Step S202, call the scoring policy interface of the first plugin to score the nodes that meet the requirements and obtain the first scoring result.

[0041] Further, step S202 includes: Calculate the first scoring result according to formula (1): (1) Wherein, WorstFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resource of the i-th node, free_mem is the remaining memory resource of the i-th node, and k1 and k2 are constant coefficients.

[0042] Further, step S202 includes: Calculate the first scoring result according to formula (2): (2) Wherein, BestFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resource of the i-th node, free_mem is the remaining memory resource of the i-th node, free_disk is the remaining disk resource of the i-th node, free_net is the remaining network resource of the i-th node, request_cpu is the CPU resource requested for task running of the i-th node, request_mem is the memory resource requested for task running of the i-th node, request_disk is the disk resource requested for task running of the i-th node, request_net is the network resource requested for task running of the i-th node, and w1, w2, w3, and w4 are constant coefficients.

[0043] Further, step S105 includes the following steps: Step S301, construct a score matrix for the scoring results corresponding to all plugins; Step S302, obtain a comprehensive score through the weighted summation algorithm for the score matrix.

[0044] Further, step S101 includes the following steps: Step S401, obtain a list of tasks to be scheduled; Step S402, traverse the jobs in the list of tasks to be scheduled according to the task priority order to obtain the job with the highest priority.

[0045] Here, if the list of tasks to be scheduled is empty, then repeat step S401.

[0046] Further, the method further includes: If all nodes do not meet the scheduling conditions, then the tasks to be scheduled are not scheduled.

[0047] This application designs a service scheduling (nginx) task and a job scheduling (pytorch job) task. Both the service scheduling task and the job scheduling task require 20% of the CPU and 20% of the memory. The job scheduling also requires an additional 20% of disk and network requirements.

[0048] The queue used by the service scheduling task is named normal, and this queue contains a Worst-Fit scheduling plugin. Worst-Fit is also called EPVM, which selects the least busy machine and implicitly distributes requests to different machines, sacrificing more fragmentation for better availability. There are four nodes A, B, C, and D in the queue. Refer to Table 1: Table 1

[0049] Among them, Node D does not meet the screening strategy, and the scoring strategy needs to score the remaining three nodes A, B, and C. The Worst-Fit strategy scoring refers to formula (1). k1 and k2 are constant coefficients, which can be determined according to the specific cluster. Here, both k1 and k2 are set to 10. Thus, it can be known that:

[0050] The result given by the Worst-Fit scheduling plugin is [19, 17.5, 18, 0]. At this time, if there are multiple scheduling plugins, the results of each plugin can be regarded as a score matrix [Score(i, A), Score(i, B), Score(i, C), Score(i, D)].

[0051] The final result can be obtained by using multi-criteria decision analysis (MCDA) to get a comprehensive score. Common algorithms include weighted summation, TOPSIS, entropy weight method, analytic hierarchy process, and principal component analysis. Here we use weighted summation, and the weight design is Wp = [1]. The result is equal to S×Wp = [19, 17.5, 18, 0], and Node A is selected for scheduling.

[0052] For the job scheduling task, the cluster administrator may have requirements for the resource utilization rate of the cluster servers, that is, when the input remains unchanged, reduce the number of machines used. The queue used by the service scheduling task is named ml, and this queue contains a Best-Fit scheduling plugin and a node priority plugin. The Best-Fit scheduling plugin can select the busiest machine in the cluster with the least fragmentation, referring to formula (2). w1, w2, w3, and w4 are constant coefficients, which can be adjusted according to actual needs, where w = 40, w1 = 10, w2 = 10, w3 = 10, w4 = 10.

[0053] The node priority plugin can artificially specify the order in which servers in the cluster are preferentially used. Combining with the Best-Fit algorithm, it can fill up the servers with better performance first.

[0054] NodeOrder_Score(A) = 10 NodeOrder_Score(B) = 10 NodeOrder_Score(C) = 20 NodeOrder_Score(D) = 20 There are four nodes A, B, C, and D in the queue. Refer to Table 2: Table 2

[0055] The Best-Fit score is [14, 15, 17, 0]. Among them, node D does not meet the scheduling constraints.

[0056] The node priority score is [10, 10, 20, 20]. Its result can be represented by a matrix S. Since node D does not meet the scheduling constraints, set the corresponding column of D to 0: [[ 14, 15, 17, 0] [ 10, 10, 20, 0]] The final result can be obtained by using multi-criteria decision analysis (MCDA) to get the comprehensive score. Here, weighted summation is used, and the weight design is Wp [0.6, 0.4]. The result is equal to S × Wp = [12.4, 13, 18.2, 0]. Select node C for scheduling. Example 2: Figure 5 It is a schematic diagram of the multi-domain distributed scheduling system of the scheduler provided in the second embodiment of the present invention.

[0057] Refer to Figure 5 This system includes: A job acquisition module for acquiring the job with the highest priority from the list of tasks to be scheduled; A queue information acquisition module for acquiring the queue information of the job with the highest priority; among them, each queue in the queue information configures multiple plugins and allocates multiple nodes; A monitoring information acquisition module for acquiring the monitoring information of each node; among them, the monitoring information includes CPU utilization, memory utilization, disk utilization, and network utilization; A calculation module for calculating the scoring results corresponding to all plugins according to the monitoring information of each node; An analysis module for analyzing the scoring results corresponding to all plugins through a multi-criteria decision analysis algorithm to obtain a comprehensive score; A selection module for selecting the node with the highest score from the comprehensive score as the scheduling node of the task to be scheduled and performing job scheduling.

[0058] The embodiments of the present invention provide a multi-domain distributed scheduling method and system for a scheduler, including: obtaining the job with the highest priority from the task list to be scheduled; obtaining the queue information of the job with the highest priority; wherein, each queue in the queue information configures multiple plugins and allocates multiple nodes; obtaining the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate, and network utilization rate; calculating the scoring results corresponding to all plugins according to the monitoring information of each node; analyzing the scoring results corresponding to all plugins through a multi-criteria decision analysis algorithm to obtain a comprehensive score; selecting the node with the highest score from the comprehensive score as the scheduling node of the task to be scheduled and performing job scheduling; which can improve the utilization rate of cluster resources in job scheduling, achieve load balancing, and meet the scheduling requirements in specific scenarios.

[0059] The embodiments of the present invention further provide an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the multi-domain distributed scheduling method of the scheduler provided in the above embodiments are implemented.

[0060] The embodiments of the present invention further provide a computer-readable medium having non-volatile program code executable by a processor. A computer program is stored on the computer-readable medium. When the computer program is run by the processor, the steps of the multi-domain distributed scheduling method of the scheduler provided in the above embodiments are executed.

[0061] The computer program product provided by the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, reference can be made to the method embodiments, which will not be elaborated here.

[0062] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system and device can refer to the corresponding processes in the foregoing method embodiments, which will not be elaborated here.

[0063] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and defined, the terms "installed", "connected", and "coupled" shall be construed in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0064] If the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0065] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0066] Finally, it should be noted that the above-described embodiments are only specific embodiments of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field of the present invention can still modify the technical solutions described in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims described.

Claims

1. A multi-domain distributed scheduling method for a scheduler, characterized by The method includes: Obtaining the job with the highest priority from the list of tasks to be scheduled; Obtaining the queue information of the job with the highest priority; wherein, multiple plugins are configured for each queue in the queue information and multiple nodes are allocated; Obtaining the monitoring information of each node; wherein, the monitoring information includes CPU utilization rate, memory utilization rate, disk utilization rate, and network utilization rate; Calculating the scoring results corresponding to all the plugins according to the monitoring information of each node; Analyzing the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; Selecting the node with the highest score from the comprehensive score as the scheduling node for the task to be scheduled, and performing job scheduling.

2. The multi-domain distributed scheduling method of the scheduler according to claim 1, characterized in that, Calculating the scoring results corresponding to all the plugins according to the monitoring information of each node includes repeatedly performing the following processing until each plugin has been traversed: Loading the first plugin and calling the filtering strategy interface of the first plugin to filter out the nodes that do not meet the requirements; Calling the scoring strategy interface of the first plugin to score the nodes that meet the requirements to obtain the first scoring result.

3. The multi-domain distributed scheduling method of the scheduler according to claim 2, wherein Calling the scoring strategy interface of the first plugin to score the nodes that meet the requirements to obtain the first scoring result, including: Calculating the first scoring result according to the following formula: wherein, WorstFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resources of the i-th node, free_mem is the remaining memory resources of the i-th node, and k1 and k2 are constant coefficients.

4. The multi-domain distributed scheduling method of the scheduler according to claim 2, characterized in that Calling the scoring strategy interface of the first plugin to score the nodes that meet the requirements to obtain the first scoring result, including: Calculating the first scoring result according to the following formula: wherein, BestFit_score(i) is the score of the i-th node, free_cpu is the remaining CPU resources of the i-th node, free_mem is the remaining memory resources of the i-th node, free_disk is the remaining disk resources of the i-th node, free_net is the remaining network resources of the i-th node, request_cpu is the CPU resources requested for task running by the i-th node, request_mem is the memory resources requested for task running by the i-th node, request_disk is the disk resources requested for task running by the i-th node, request_net is the network resources requested for task running by the i-th node, and w1, w2, w3, and w4 are constant coefficients.

5. The multi-domain distributed scheduling method of the scheduler according to claim 1, characterized in that Analyzing the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score, including: Constructing the scoring results corresponding to all the plugins into a score matrix; Obtaining the comprehensive score by the weighted summation algorithm for the score matrix.

6. The multi-domain distributed scheduling method of the scheduler according to claim 1, characterized in that, Obtaining the job with the highest priority from the list of tasks to be scheduled, including: Obtaining the list of tasks to be scheduled; Traversing the jobs in the list of tasks to be scheduled according to the task priority order to obtain the job with the highest priority.

7. The multi-domain distributed scheduling method of the scheduler according to claim 1, characterized in that The method further includes: If all the nodes do not meet the scheduling conditions, the task to be scheduled is not scheduled.

8. A multi-domain distributed scheduling system for a scheduler, characterized in that, The system includes: A job acquisition module, configured to acquire the job with the highest priority from the list of tasks to be scheduled; A queue information acquisition module, configured to acquire the queue information of the job with the highest priority; wherein, each queue in the queue information configures multiple plugins and allocates multiple nodes; A monitoring information acquisition module, configured to acquire the monitoring information of each node; wherein, the monitoring information includes CPU utilization, memory utilization, disk utilization, and network utilization; A calculation module, configured to calculate the scoring results corresponding to all the plugins according to the monitoring information of each node; An analysis module, configured to analyze the scoring results corresponding to all the plugins through a multi-index decision analysis algorithm to obtain a comprehensive score; A selection module, configured to select the node with the highest score from the comprehensive score as the scheduling node of the task to be scheduled and execute job scheduling.

9. An electronic device, comprising a memory and a processor, wherein a computer program capable of running on the processor is stored on the memory, characterized in that When the processor executes the computer program, it implements the method according to any one of claims 1 to 7 above.

10. A computer-readable medium having non-volatile program code executable by a processor, characterized in that, The program code causes the processor to execute the method according to any one of claims 1 to 7.

Images