CN117632505A - Heterogeneous computing power intelligent scheduling system and method - Google Patents

Abstract

The invention belongs to the technical field of computing power task scheduling, and discloses a heterogeneous computing power intelligent scheduling system and method. The heterogeneous computing power intelligent scheduling system includes a computing power scheduling module, a cloud native storage ETCD module, and a third-party expansion strategy module. The cloud native storage ETCD module stores heterogeneous computing power resource CRD, scheduling strategy CRD, computing task CRD, and cluster CRD. ; The third-party expansion policy module is used to register the third-party expansion policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD; the cloud native storage ETCD module is used to update all the data according to the third-party expansion policy The scheduling policy CRD is used to obtain the updated scheduling policy CRD; the computing power scheduling module is used to determine the target cluster in the cluster CRD according to the updated scheduling policy CRD. It can realize the compatibility and expansion functions of heterogeneous computing power scheduling.

Description

Heterogeneous computing power intelligent scheduling system and method

Technical field

The present invention relates to the technical field of computing power task scheduling, and in particular to a heterogeneous computing power intelligent scheduling system and method.

Background technique

With the deepening of digital transformation in production operations of various industries, various industry terminals will generate massive amounts of raw data, so a large amount of computing power is urgently needed for processing. In order to meet the requirements of second-level response, low latency and business continuity at the business layer , often using computing power clusters to provide high availability, high throughput and high scalability technical architecture. In the process of building computing power clusters year by year, there are a large number of heterogeneous computing power cluster resources in the same data center or different data centers. How can these computing power resources be managed and scheduled on a unified platform so that users of computing power can conveniently Allocating computing resources according to needs and usage habits is a problem that needs to be solved. For intelligent scheduling of heterogeneous computing power, the current mainstream technical solutions are to customize a set of intelligent scheduling methods according to specific scenarios. Each scheduling solution cannot match the needs of heterogeneous computing power scheduling in other scenarios, and existing scheduling The solution cannot provide scalability. It can be seen that the existing scheduling solution lacks compatibility and scalability.

Contents of the invention

The main purpose of the present invention is to provide a heterogeneous computing power intelligent scheduling system and method, aiming to solve the technical problem of lack of compatibility and scalability of scheduling solutions in the prior art.

In order to achieve the above objectives, the present invention provides a heterogeneous computing power intelligent scheduling system. The heterogeneous computing power intelligent scheduling system includes a computing power scheduling module, a cloud native storage ETCD module and a third-party expansion strategy module. The cloud native The storage ETCD module stores heterogeneous computing resource CRD, scheduling strategy CRD, computing task CRD and cluster CRD. The unified model and interface library provided by the heterogeneous computing resource CRD are implemented in the heterogeneous computing intelligent scheduling system. The scheduling policy CRD is built in; where,

The third-party expansion policy module is used to register the third-party expansion policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD;

The cloud native storage ETCD module is used to update the scheduling policy CRD according to the third-party extension policy to obtain the updated scheduling policy CRD;

The computing power scheduling module is used to determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, and bind the computing power task to the target cluster.

Optionally, the heterogeneous computing resource CRD is defined based on cloud native technology. The heterogeneous computing resource CRD includes an interface abstraction layer. The interface abstraction layer uniformly handles unit adaptation of computing resources.

Optionally, the interface abstraction layer includes a Deep Copy deep copy interface, a Canonicalize hexadecimal normalization conversion interface, an ADD addition operation interface, a SUB subtraction operation interface, a CMP comparison operation interface, and an Equal equality judgment operation interface.

Optionally, the heterogeneous computing resource CRD provides a unified addition, deletion, modification and query interface through k8s apiserver.

Optionally, the updated scheduling strategy CRD includes a built-in strategy and a third-party expansion strategy. The built-in strategy is loaded into the Cache during the running phase of the computing power scheduling module, and the third-party expansion strategy is loaded into the Cache during the running phase of the computing power scheduling module. The running phase of the module is read from the cloud native storage ETCD module through dynamic loading.

In addition, to achieve the above objectives, the present invention also proposes an intelligent scheduling method for heterogeneous computing power. The intelligent scheduling method for heterogeneous computing power includes:

Register the third-party expansion strategy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD;

Update the scheduling policy CRD in the cloud native storage ETCD module according to the third-party extension policy to obtain the updated scheduling policy CRD;

Determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, and bind the computing power task to the target cluster.

Optionally, the update scheduling policy CRD includes a built-in strategy and a third-party extension policy; wherein the target cluster in the cluster CRD is determined according to the update scheduling policy CRD, and the computing power task is bound to the cluster CRD. Target clusters include:

After the computing power task is created according to the computing power task CRD in the cloud native storage ETCD module, the computing power scheduling module is triggered to start the scheduling workflow, wherein the scheduling workflow includes a pre-selection stage, a optimization stage and a binding stage;

In the pre-selection stage, the built-in policies are traversed, and the first cluster that satisfies all built-in policies is selected from the cluster CRD;

In the optimization stage, filter out a second cluster from the first cluster according to the third-party expansion strategy;

In the binding stage, a target cluster in the second cluster is determined, and the computing power task is bound to the target cluster.

Optionally, selecting the second cluster from the first cluster according to the third-party expansion policy includes:

S301. Call the third-party Operator/schedule HTTPS interface to execute the target policy in the third-party expansion policy, and score the clusters in the first cluster that meet the target policy to obtain the scheduling result;

S302. Determine whether there is a new target policy in the third-party expansion policy;

S303. If yes, use the new target policy as the target policy in step S301 to repeat steps S301-S303; if not, filter out the second cluster from the first cluster based on the scheduling result.

Optionally, selecting the second cluster from the first cluster based on the scheduling result includes:

Set preset threshold;

Determine the weight score of each cluster in the first cluster according to the scheduling result;

The cluster with a weight score greater than the preset threshold in the first cluster is regarded as the second cluster.

Optionally determining a target cluster in the second cluster and binding the computing power task to the target cluster includes:

Determine the weight score of each cluster in the second cluster according to the scheduling result;

The cluster with the highest weight score in the second cluster is used as the target cluster, and the computing power task is bound to the target cluster.

The invention proposes a heterogeneous computing power intelligent scheduling system and method. The heterogeneous computing power intelligent scheduling system includes a computing power scheduling module, a cloud native storage ETCD module and a third-party expansion strategy module. The cloud native storage ETCD module stores Heterogeneous computing resource CRD, scheduling strategy CRD, computing task CRD and cluster CRD. The unified model and interface library provided by the heterogeneous computing resource CRD implement the scheduling built into the heterogeneous computing intelligent scheduling system. Policy CRD; wherein, the third-party extension policy module is used to register the third-party extension policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD; the cloud native storage The ETCD module is used to update the scheduling policy CRD according to the third-party extension policy to obtain the updated scheduling policy CRD; the computing power scheduling module is used to determine the target cluster in the cluster CRD according to the updated scheduling policy CRD. , and bind the computing power task to the target cluster. Through the above system, a unified framework for heterogeneous computing power intelligent scheduling is used, combined with the Operator, CRD and built-in strategies of Kubernetes to achieve the compatibility and expansion functions of heterogeneous computing power scheduling strategies.

Description of drawings

Figure 1 is a structural block diagram of the first embodiment of the heterogeneous computing power intelligent scheduling system of the present invention;

Figure 2 is a schematic diagram of heterogeneous computing power scheduling in the first embodiment of the heterogeneous computing power intelligent scheduling system of the present invention;

Figure 3 is a schematic flow chart of the first embodiment of the intelligent scheduling method for heterogeneous computing power of the present invention;

Figure 4 is a schematic flowchart of determining the second cluster in the first embodiment of the heterogeneous computing power intelligent scheduling method of the present invention;

Figure 5 is a scheduling flow chart in the first embodiment of the intelligent scheduling method for heterogeneous computing power of the present invention.

The realization of the purpose, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

It should be noted that ETCD represents a distributed key-value storage system, and CRD (Custom ResourceDefinition) is a mechanism in Kubernetes that is used to extend the Kubernetes API and allow users to define custom resources; Operator Controller is a custom controller , used to manage and automate specific types of applications or resources in a Kubernetes cluster.

The present invention proposes an intelligent scheduling system for heterogeneous computing power.

Referring to Figure 1, in the embodiment of the present invention, the heterogeneous computing power intelligent scheduling system includes a computing power scheduling module 10, a cloud native storage ETCD module 20 and a third-party expansion strategy module 30. The cloud native storage ETCD module 20 stores There are heterogeneous computing resource CRD, scheduling policy CRD, computing task CRD and cluster CRD. The unified model and interface library provided by the heterogeneous computing resource CRD are implemented in the heterogeneous computing intelligent scheduling system. Scheduling policy CRD; wherein, the third-party expansion policy module 30 is used to register the third-party expansion policy to the cloud native storage ETCD module 20 through the computing task interface provided by the heterogeneous computing resource CRD; the The cloud native storage ETCD module 20 is used to update the scheduling policy CRD according to the third-party expansion policy to obtain an updated scheduling policy CRD; the computing power scheduling module 10 is used to determine the cluster according to the updated scheduling policy CRD Target cluster in CRD, and bind computing tasks to the target cluster.

It should be noted that the scheduling policy includes built-in policies and third-party extension policies. The scheduling policy is stored in the cloud native storage ETCD module through the CRD data mode; the updated scheduling policy CRD refers to the updated scheduling policy CRD; the computing power task passes the CRD The data pattern is stored in the cloud native storage ETCD module; after multi-cluster information is reported, it is stored in the cloud native storage ETCD module through the CRD data pattern; the computing power scheduling module refers to the computing power scheduling Operator Controller; the third-party expansion strategy module refers to Third-party expansion policy Operator Controller. The third-party expansion policy Operator Controller needs to implement the /schedule HTTP interface. The third-party expansion policy module is mainly used to register the third-party expansion policy into the scheduling policy CRD of the cloud native storage ETCD module;

It is understandable that the heterogeneous computing power intelligent scheduling system can register at any time to trigger dynamic loads during operation; the target cluster can be the cluster with the highest degree of compliance in the cluster CRD, and can be used as the cluster that needs to be run for this computing power task.

It should be noted that updating the scheduling policy CRD according to the third-party extension policy is dynamically updated in real time and can be implemented based on the asynchronous watch mechanism. Specifically, when a third-party user updates the scheduling policy CRD based on the third-party extension policy module, The following two steps are implemented: 1. Based on the unified model and interface library provided by heterogeneous computing resources CRD, third parties can implement extended computing power scheduling Operator Controller and provide a standard HTTPS computing power task scheduling interface (/schedule). HTTPS can ensure communication and data security; 2. Register the third-party extension policy to the cloud native storage ETCD, triggering the policy Cache update of the computing power scheduling Operator Controller. Register not only the extended policy, but also the callback access address, so that Let the computing power scheduling Operator Controller know the address of the /schedule interface call.

In one embodiment, the heterogeneous computing resource CRD is defined based on cloud native technology. The heterogeneous computing resource CRD includes an interface abstraction layer. The interface abstraction layer uniformly handles unit adaptation of computing resources.

It should be noted that the interface abstraction layer uniformly handles the unit adaptation of binary, decimal, floating point and other computing resources; the interface abstraction layer includes the adaptation of all known computing resource units.

In this embodiment, all heterogeneous computing power resources are compatible through the support of different bases and units, and the problem of a unified model and measurement of heterogeneous computing power is solved.

In one embodiment, the interface abstraction layer includes a Deep Copy deep copy interface, a Canonicalize hexadecimal normalization conversion interface, an ADD addition operation interface, a SUB subtraction operation interface, a CMP comparison operation interface, and an Equal equality judgment operation interface.

In this embodiment, through the Deep Copy deep copy interface, Canonicalize hexadecimal normalization conversion interface, ADD addition operation interface, SUB subtraction operation interface, CMP comparison size operation interface and Equal equality judgment operation interface in the interface abstraction layer, the problem can be solved Issues such as conversion, addition, subtraction, and comparison of heterogeneous computing resources.

In one embodiment, the heterogeneous computing resource CRD provides a unified addition, deletion, modification and query interface through k8s apiserver.

In one embodiment, the updated scheduling policy CRD includes a built-in policy and a third-party expansion policy. The built-in policy is loaded into the Cache during the running phase of the computing power scheduling module, and the third-party expansion policy is loaded into the Cache during the running phase of the computing power scheduling module. The running phase of the force scheduling module is read from the cloud native storage ETCD module through dynamic loading.

In the specific implementation, as shown in the schematic diagram of heterogeneous computing power scheduling in Figure 2, it can be determined that the heterogeneous computing power intelligent scheduling system includes computing power scheduling Operator Controller, cloud native storage ETCD and third-party expansion strategy Operator Controller. Among them, cloud native Storage ETCD stores heterogeneous computing resource CRD, scheduling policy CRD, computing task CRD and cluster CRD. The unified interface abstraction layer of heterogeneous computing resource CRD includes Deep Copy deep copy interface, Canonicalize hexadecimal standardization conversion interface, ADD addition Operation interface, SUB subtraction operation interface, CMP comparison operation interface and Equal equality judgment operation interface.

It should be noted that the reference model of heterogeneous computing resource CRD is a YAML configuration file of a custom resource definition (CustomResourceDefinition) object in Kubernetes, which defines a custom resource named "kuberesources.resource.kubearena.io" Resource; the API version of this custom resource is "apiextensions.k8s.io/v1" and belongs to the "resource.kubearena.io" group. Its object type is "KubeResource", the plural form is "KubeResourceList", and has a short name "res"; this custom resource is accessible throughout the cluster and is provided with a version named "v1"; this version The definition contains specification and state information about the "KubeResource" object.

In this embodiment, the heterogeneous computing power intelligent scheduling system includes a computing power scheduling module, a cloud native storage ETCD module, and a third-party expansion strategy module. The cloud native storage ETCD module stores heterogeneous computing power resources CRD and scheduling strategy CRD. , computing power task CRD and cluster CRD. The unified model and interface library provided by the heterogeneous computing power resource CRD implement the scheduling policy CRD built into the heterogeneous computing power intelligent scheduling system; wherein, the third-party extension The policy module is used to register the third-party expansion policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD; the cloud native storage ETCD module is used to register the third-party expansion policy according to the third-party The expansion policy updates the scheduling policy CRD to obtain an updated scheduling policy CRD; the computing power scheduling module is used to determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, and bind the computing power task to the cluster CRD. Describe the target cluster. Through the above system, a unified framework for heterogeneous computing power intelligent scheduling is used, combined with the Operator, CRD and built-in strategies of Kubernetes to achieve the compatibility and expansion functions of heterogeneous computing power scheduling strategies.

In addition, as shown in Figure 3, embodiments of the present invention also provide an intelligent scheduling method for heterogeneous computing power.

In this embodiment, the heterogeneous computing power intelligent scheduling method includes the following steps:

Step S10: Register the third-party expansion policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD.

It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a mobile phone, a tablet computer, a personal computer, etc., or an electronic device capable of realizing the above functions. Equipment or heterogeneous computing power intelligent scheduling system. The following takes the heterogeneous computing power intelligent scheduling system as an example to describe this embodiment and the following embodiments.

It should be noted that the computing task interface provided by the heterogeneous computing resource CRD refers to the standard HTTPS computing task scheduling interface (/schedule);

Step S20: Update the scheduling policy CRD in the cloud native storage ETCD module according to the third-party extension policy to obtain the updated scheduling policy CRD.

It is understandable that the heterogeneous computing power intelligent scheduling system can register at any time to trigger dynamic loads during operation; the target cluster can be the cluster with the highest degree of compliance in the cluster CRD, and can be used as the cluster that needs to be run for this computing power task.

It should be noted that updating the scheduling policy CRD refers to the updated scheduling policy CRD; updating the scheduling policy CRD includes built-in policies and third-party extension policies; updating the scheduling policy CRD according to the third-party extension policy is a real-time dynamic update It can be implemented based on the asynchronous watch mechanism. Specifically, when third-party users update the scheduling policy CRD based on the third-party extended policy module, it can be implemented through the following two steps: 1. Based on the unified model and interface library provided by the heterogeneous computing resource CRD To enable third parties to implement extended computing power scheduling Operator Controller, and provide a standard HTTPS computing power task scheduling interface (/schedule), HTTPS can ensure communication and data security; 2. Register the third-party extension policy to the cloud native storage ETCD, trigger When updating the policy Cache of the computing power scheduling Operator Controller, not only the extended policy is registered, but also the callback access address is registered, so that the computing power scheduling Operator Controller knows the address of the /schedule interface call.

Step S30: Determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, bind the computing power task to the target cluster, and bind the computing power task to the target cluster.

It should be noted that after the computing task is bound to the target cluster and the current computing task is scheduled, you can continue to create and schedule new computing tasks based on the computing task CRD stored in the cloud native storage ETCD module.

In one embodiment, the updated scheduling policy CRD includes a built-in policy and a third-party extension policy; wherein the target cluster in the cluster CRD is determined according to the updated scheduling policy CRD, and the computing task is bound to The target cluster includes:

After the computing power task is created according to the computing power task CRD in the cloud native storage ETCD module, the computing power scheduling module is triggered to start the scheduling workflow, wherein the scheduling workflow includes a pre-selection stage, a optimization stage and a binding stage;

In the pre-selection stage, the built-in policies are traversed, and the first cluster that satisfies all built-in policies is selected from the cluster CRD;

In the optimization stage, filter out a second cluster from the first cluster according to the third-party expansion strategy;

In the binding stage, a target cluster in the second cluster is determined, and the computing power task is bound to the target cluster.

It should be noted that the scheduling workflow Workflow includes three stages, which are the pre-selection stage, the priority stage and the binding stage. In the pre-selection stage, all built-in strategies need to be traversed and executed. For a single computing task, the built-in strategies are configured according to the configuration. are executed in sequence (executing built-in policies can be understood as judging whether the cluster meets the built-in policies); the first cluster refers to the cluster in the cluster CRD that satisfies all built-in policies; in the optimization stage, third-party expansion strategies need to be executed (which can be understood as Determine whether the cluster meets the third-party expansion policy), the second cluster refers to the cluster in the first cluster that meets the third expansion policy; in the binding phase, the cluster with the highest compliance in the second cluster can be used as the target cluster, and the target cluster That is the cluster that needs to be run for this computing task.

It should be noted that when executing a third-party expansion strategy, you need to call the third-party Operator/scheduleHTTPS interface and receive the scheduling results; the parameters of the /schedule interface are the computing task (Compute Task) and the previously filtered cluster (cluster CRD) list , returns the cluster list filtered out this time (i.e. the second cluster).

In this embodiment, the target cluster is determined from the cluster CRD through the three major stages of the scheduling workflow, which not only can quickly determine the cluster with the highest degree of compliance, but can also effectively improve the accuracy of scheduling.

In one embodiment, as shown in Figure 4, selecting the second cluster from the first cluster according to the third-party expansion policy includes:

Step S301: Call the third-party Operator/schedule HTTPS interface to execute the target policy in the third-party extension policy, and score the clusters in the first cluster that meet the target policy to obtain the scheduling result.

Step S302: Determine whether there is a new target policy in the third-party extension policy.

Step S303: If yes, use the new target policy as the target policy in step S301 to repeatedly execute steps S301-S303; if not, filter out the second cluster from the first cluster based on the scheduling result.

It is understood that the new target policy can be understood as the next third-party expansion policy of the target policy in the third-party expansion policy.

It should be noted that the target strategy can be determined according to the registration order of the third-party expansion strategy. For example, when performing the first round of cluster scoring, the target strategy is the first registered strategy among the third-party expansion strategies; for each round of clustering The scheduling results need to be updated for the scoring of The scheduling result is: cluster A gets 0 points and cluster B gets 1 point; during the second round of cluster scoring, since both cluster A and cluster B meet the target strategy in the second round of scoring, after the second round of scoring The updated scheduling result is: cluster A gets 1 point, and cluster B gets 2 points.

It can be understood that when there is no new target policy in the third-party expansion policy, the scheduling result obtained is the final scheduling result.

In one embodiment, selecting the second cluster from the first cluster based on the scheduling result includes:

Set preset threshold;

Determine the weight score of each cluster in the first cluster according to the scheduling result;

The cluster with a weight score greater than the preset threshold in the first cluster is regarded as the second cluster.

It should be noted that the scheduling result refers to the final scheduling result, which can be used to determine the weight score of each cluster in the first cluster; the preset threshold can be set in advance, and preferably, the preset threshold can be set to 0 , since the weight score of the cluster is greater than or equal to 0, when the weight score of the cluster is not 0, the cluster can be used as the second cluster. The cluster with a weight score of 0 can be regarded as all target strategies in the third-party expansion strategy. Clusters that are not satisfied.

In this embodiment, by setting a threshold to determine the second cluster based on the scores of each cluster in the first cluster, the efficiency of determining the second cluster can be effectively improved.

In one embodiment, determining the target cluster in the second cluster and binding the computing power task to the target cluster includes:

Determine the weight score of each cluster in the second cluster according to the scheduling result;

The cluster with the highest weight score in the second cluster is used as the target cluster, and the computing power task is bound to the target cluster.

It should be noted that the scheduling result refers to the final scheduling result, which can be used to determine the weight score of each cluster in the second cluster; specifically, for example, the weight score of each cluster in the second cluster is determined based on the scheduling result as: If the weight of cluster A is 1 point and the weight of cluster B is 2, then cluster B is the target cluster in the second cluster.

It should be noted that after the computing task is bound to the target cluster and the current computing task is scheduled, you can continue to create and schedule new computing tasks based on the computing task CRD stored in the cloud native storage ETCD module.

In this embodiment, by selecting the cluster with the highest weight score as the cluster that needs to be run for this computing power task, the accuracy of computing power scheduling can be effectively improved.

In the specific implementation, as shown in Figure 5, it is necessary to define heterogeneous computing power resource CRD based on cloud native technology, and implement a unified interface abstraction layer through heterogeneous computing power resource CRD to complete the unified model and measurement of heterogeneous computing power. framework, and then implement the built-in scheduling strategy based on the unified model and interface library provided by the heterogeneous computing resource CRD. Then the computing task can be created to trigger the scheduling workflow, and the computing task can be completed based on the three stages of the scheduling workflow. of scheduling.

It should be understood that the above are only examples and do not constitute any limitation on the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any limitations on this.

It should be noted that the workflow described above is only illustrative and does not limit the scope of the present invention. In practical applications, those skilled in the art can select some or all of them for implementation according to actual needs. The purpose of this embodiment is not limited here.

In addition, for technical details that are not described in detail in this embodiment, please refer to the intelligent scheduling method for heterogeneous computing power provided by any embodiment of the present invention, and will not be described again here.

Furthermore, it should be noted that, as used herein, the terms "include", "comprises" or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or system that includes a list of elements includes not only those elements, but also other elements not expressly listed or elements inherent to the process, method, article or system. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

The above serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. implementation. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product that is essentially or contributes to the existing technology. The computer software product is stored in a storage medium (such as a read-only memory). , ROM)/RAM, magnetic disk, optical disk), including several instructions to cause a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the method described in various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the description and drawings of the present invention may be directly or indirectly used in other related technical fields. , are all similarly included in the scope of patent protection of the present invention.

Claims (10)

1. A heterogeneous computing power intelligent scheduling system, characterized in that the heterogeneous computing power intelligent scheduling system includes a computing power scheduling module, a cloud native storage ETCD module and a third-party expansion strategy module. The cloud native storage ETCD module Heterogeneous computing resource CRD, scheduling policy CRD, computing task CRD and cluster CRD are stored. The unified model and interface library provided by the heterogeneous computing resource CRD are implemented in the heterogeneous computing intelligent scheduling system. Described scheduling policy CRD; where, The third-party expansion policy module is used to register the third-party expansion policy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD; The cloud native storage ETCD module is used to update the scheduling policy CRD according to the third-party extension policy to obtain the updated scheduling policy CRD; The computing power scheduling module is used to determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, and bind the computing power task to the target cluster. 2. The system of claim 1, wherein the heterogeneous computing resource CRD is defined based on cloud native technology, the heterogeneous computing resource CRD includes an interface abstraction layer, and the interface abstraction layer processes computing in a unified manner. Unit adaptation of human resources. 3. The system of claim 2, wherein the interface abstraction layer includes a Deep Copy deep copy interface, a Canonicalize hexadecimal normalization conversion interface, an ADD addition operation interface, a SUB subtraction operation interface, a CMP comparison operation interface, and Equal equality judgment operation interface. 4. The system of claim 1, wherein the heterogeneous computing resource CRD provides a unified interface for adding, deleting, modifying, and checking through k8s apiserver. 5. The system of claim 1, wherein the updated scheduling strategy CRD includes a built-in strategy and a third-party extension strategy, and the built-in strategy is loaded into the Cache during the running phase of the computing power scheduling module, so The third-party expansion strategy is read from the cloud native storage ETCD module through dynamic loading during the running phase of the computing power scheduling module. 6. An intelligent scheduling method for heterogeneous computing power, characterized in that the intelligent scheduling method for heterogeneous computing power is applied to the intelligent scheduling system for heterogeneous computing power according to any one of claims 1-5, and the Intelligent scheduling methods for heterogeneous computing power include: Register the third-party expansion strategy to the cloud native storage ETCD module through the computing task interface provided by the heterogeneous computing resource CRD; Update the scheduling policy CRD in the cloud native storage ETCD module according to the third-party extension policy to obtain the updated scheduling policy CRD; Determine the target cluster in the cluster CRD according to the updated scheduling policy CRD, and bind the computing power task to the target cluster. 7. The method of claim 6, wherein the update scheduling policy CRD includes a built-in policy and a third-party extension policy; wherein the target cluster in the cluster CRD is determined according to the update scheduling policy CRD. , and bind computing tasks to the target cluster, including: After the computing power task is created according to the computing power task CRD in the cloud native storage ETCD module, the computing power scheduling module is triggered to start the scheduling workflow, wherein the scheduling workflow includes a pre-selection stage, a optimization stage and a binding stage; In the pre-selection stage, the built-in policies are traversed, and the first cluster that satisfies all built-in policies is selected from the cluster CRD; In the optimization stage, filter out a second cluster from the first cluster according to the third-party expansion strategy; In the binding stage, a target cluster in the second cluster is determined, and the computing power task is bound to the target cluster. 8. The method of claim 7, wherein filtering out the second cluster from the first cluster according to the third-party expansion policy includes: S301. Call the third-party Operator/schedule HTTPS interface to execute the target policy in the third-party expansion policy, and score the clusters in the first cluster that meet the target policy to obtain the scheduling result; S302. Determine whether there is a new target policy in the third-party expansion policy; S303. If yes, use the new target policy as the target policy in step S301 to repeat steps S301-S303; if not, filter out the second cluster from the first cluster based on the scheduling result. 9. The method of claim 8, wherein filtering out the second cluster from the first cluster based on the scheduling result includes: Set preset threshold; Determine the weight score of each cluster in the first cluster according to the scheduling result; The cluster with a weight score greater than the preset threshold in the first cluster is regarded as the second cluster. 10. The method of claim 8, wherein determining the target cluster in the second cluster and binding the computing power task to the target cluster includes: Determine the weight score of each cluster in the second cluster according to the scheduling result; The cluster with the highest weight score in the second cluster is used as the target cluster, and the computing power task is bound to the target cluster.