WO2024098872A1

WO2024098872A1 - Task processing method, system, apparatus and device, and computer-readable storage medium

Info

Publication number: WO2024098872A1
Application number: PCT/CN2023/113135
Authority: WO
Inventors: 张亚强; 李茹杨; 邓琪; 李雪雷; 赵雅倩; 李仁刚
Original assignee: 山东海量信息技术研究院
Priority date: 2022-11-07
Filing date: 2023-08-15
Publication date: 2024-05-16
Also published as: CN115421930A; CN115421930B

Abstract

A task processing method, system, apparatus and device, and a computer-readable storage medium, which are applied to the technical field of resource scheduling. The method is applied to computing nodes in an edge network, wherein each computing node is provided with one or more heterogeneous resources. The method comprises: receiving a resource allocation policy, which is issued by an edge network, wherein the resource allocation policy comprises processing amounts regarding a target task that are allocated to the respective heterogeneous resources; inputting the resource allocation policy and local resource information into a preset policy learning model for processing, so as to obtain a resource sharing policy, wherein the resource sharing policy comprises available resources of the heterogeneous resources, and the preset policy learning model is obtained by means of performing training with task processing values of the computing nodes as an optimization target; and executing the target task by using the resource sharing policy. By means of the method, more rational resource allocation can be performed on a computing task, and the win-win of a task requester and a resource provider is realized.

Description

Task processing method, system, device, equipment and computer readable storage medium

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to the Chinese patent application filed with the China Patent Office on November 7, 2022, with application number 202211381753.2, and application name “Task processing method, system, device, equipment and computer-readable storage medium”, all contents of which are incorporated by reference in this application.

Technical Field

The present application relates to the technical field of resource scheduling, and in particular to a task processing method, system, device, equipment and computer-readable storage medium.

Background technique

Edge computing, as an emerging computing paradigm in recent years, effectively reduces network latency by providing computing power, storage, caching and other services to end users at the edge of mobile networks. Due to the variety of edge computing nodes, such as servers in operator base stations and personal computers in smart homes, computing resource nodes in edge networks have heterogeneous characteristics, and the performance of each node varies greatly.

Heterogeneous computing is a computing system composed of computing units using different types of instruction sets and architectures. The computing resources in heterogeneous computing include CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array) and ASIC (Application Specific Integrated Circuit). These computing resources each have distinct functional characteristics, different characteristics in computing power and power consumption, and can adapt to different types of computing tasks. In edge computing, different edge nodes have heterogeneous characteristics, and the computing resources inside these heterogeneous edge nodes often also have heterogeneous characteristics. In a node, multiple heterogeneous computing resources will be included at the same time.

The existing research and methods of resource allocation in edge networks mainly use resource utilization efficiency and user experience quality as the main indicators to measure the advantages and disadvantages of resource allocation strategies, and then determine the optimal solution for resource allocation. However, its main shortcoming is that it does not take into account the demand for benefits of computing node owners in edge networks. In actual scenarios, each edge node usually belongs to different owners, and the purpose of sharing its own computing resources is mainly to use idle computing resources to obtain corresponding benefits and achieve a win-win situation. At the same time, existing work pays little attention to the resource allocation problem of edge nodes with multiple heterogeneous computing resources, and the research on the collaborative processing of computing tasks based on heterogeneous resources is not sufficient.

Therefore, how to make more reasonable resource allocation for computing tasks and achieve a win-win situation for task requesters and resource providers is an urgent problem to be solved by those skilled in the art.

Summary of the invention

The purpose of this application is to provide a task processing method, which can make more reasonable resource allocation for computing tasks and achieve a win-win situation for task requesters and resource providers; another purpose of this application is to provide another task processing method, task processing device and equipment, and computer-readable storage medium, all of which have the above-mentioned beneficial effects.

In the first aspect, the present application provides a task processing method, which is applied to computing nodes in an edge network. The node is provided with one or more heterogeneous resources, and the method includes:

Receive the resource allocation strategy issued by the edge network; the resource allocation strategy includes the processing capacity of each heterogeneous resource allocated to the target task;

The resource allocation strategy and local resource information are input into the preset strategy learning model for processing to obtain the resource sharing strategy; the resource sharing strategy includes the available resources of various heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

Utilize resource sharing strategies to execute target tasks.

In some embodiments, after executing the target task using the resource sharing strategy, the method further includes:

Get the execution result of the target task;

Report the execution results to the edge network.

In some embodiments, heterogeneous resources include CPU resources, GPU resources, and FPGA resources.

In some embodiments, the task processing method further includes:

Real-time monitoring of local resource information;

When the local resource information meets the preset conditions, the local resource information is published to the edge network.

In some embodiments, when the local resource information meets a preset condition, publishing the local resource information to the edge network includes:

Determine available resources based on local resource information;

When the resource ratio of available resources reaches a preset threshold, the local resource information is published to the edge network.

In some embodiments, the local resource information includes available resources and resource provision time, and the available resources include currently available resources of various heterogeneous resources.

In a second aspect, the present application provides another task processing method, which is applied to an edge network, where each computing node in the edge network is provided with one or more heterogeneous resources, and the method includes:

Determine the target computing node and resource allocation strategy according to the task information of the target task and the local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The resource allocation strategy is sent to each target computing node so that each target computing node uses a preset strategy learning model to process the resource allocation strategy and local resource information, obtains a resource sharing strategy, and uses the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each heterogeneous resource, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

In some embodiments, determining the target computing node according to the task information of the target task and the local resource information of each computing node includes:

Determine resource provision time and available resources based on local resource information;

When both the resource provision time and the available resources meet the task requirements indicated by the task information, the computing node is determined to be the target computing node.

In some embodiments, before determining the target computing node and the resource allocation strategy according to the task information of the target task and the local resource information of each computing node, the method further includes:

The local resource information published by each computing node is counted; wherein the local resource information is published to the edge network by the computing node when the local resource information meets the preset conditions.

In some embodiments, the task processing method further includes:

Get the execution result of the target task uploaded by the target computing node;

Feedback the execution results to the initiator of the target task.

In a third aspect, the present application further discloses a task processing system, including an edge network and computing nodes deployed in the edge network, each computing node is provided with one or more heterogeneous resources, wherein:

The edge network is used to determine the target computing node and the resource allocation strategy according to the task information of the target task and the local resource information of each computing node, and send the resource allocation strategy to each target computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The target computing node is used to process the resource allocation strategy and local resource information using the preset strategy learning model, obtain the resource sharing strategy, and use the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of various heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

In a fourth aspect, the present application further discloses a task processing device, which is applied to a computing node in an edge network, each computing node is provided with one or more heterogeneous resources, and the device includes:

A receiving module is used to receive a resource allocation strategy issued by an edge network; the resource allocation strategy includes the processing capacity of each heterogeneous resource allocated to a target task;

A processing module is used to input the resource allocation strategy and local resource information into a preset strategy learning model for processing to obtain a resource sharing strategy; the resource sharing strategy includes the available resources of various heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

The execution module is used to execute the target task using the resource sharing strategy.

In a fifth aspect, the present application further discloses another task processing device, which is applied to an edge network, wherein each computing node in the edge network is provided with one or more heterogeneous resources, and the device comprises:

A determination module, used to determine the target computing node and resource allocation strategy according to the task information of the target task and the local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The sending module is used to send the resource allocation strategy to each target computing node, so that each target computing node uses the preset strategy learning model to process the resource allocation strategy and local resource information, obtains the resource sharing strategy, and uses the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each heterogeneous resource, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

In a sixth aspect, the present application further discloses a task processing device, including:

Memory for storing computer programs;

A processor is used to implement the steps of any of the above task processing methods when executing a computer program.

In a seventh aspect, the present application further discloses a non-volatile computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above task processing methods are implemented.

By applying the technical solution provided in the present application, a policy learning model is pre-deployed in each computing node of the edge network, and the resource sharing strategy is determined based on the preset policy learning model, so as to realize task processing by using the resource sharing strategy. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the obtained resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and then the computing node learns the resource allocation strategy based on the resource allocation strategy, and both the resource allocation strategy and the resource sharing strategy take the heterogeneous resources in the computing node into consideration, thereby further realizing the reasonable allocation of heterogeneous resources within the computing node.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the prior art and the present application in some embodiments, the following briefly introduces the drawings required to be used in the description of the prior art and the present application in some embodiments. Of course, the following description of the drawings in some embodiments of the present application is only a part of the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without creative work, and the obtained other drawings also belong to the protection scope of the present application.

FIG1 is a schematic diagram of the structure of a task processing system provided in some embodiments of the present application;

FIG2 is a flowchart of a task processing method provided in some embodiments of the present application;

FIG3 is a flowchart of another task processing method provided in some embodiments of the present application;

FIG4 is a flowchart of another task processing method provided in some embodiments of the present application;

FIG5 is a schematic diagram of a process flow of a task processing device provided by the present application in some embodiments;

FIG6 is a flowchart of another task processing device provided in some embodiments of the present application;

FIG7 is a schematic diagram of the structure of a task processing device provided in some embodiments of the present application;

FIG8 is a schematic diagram of the structure of a non-volatile computer-readable storage medium provided in some embodiments of the present application.

Detailed ways

The core of this application is to provide a task processing method, which can make more reasonable resource allocation for computing tasks and achieve a win-win situation for task requesters and resource providers; another core of this application is to provide another task processing method, task processing device and equipment, and computer-readable storage medium, all of which have the above-mentioned beneficial effects.

In order to describe the technical solutions in the present application more clearly and completely, the technical solutions in some embodiments of the present application will be introduced below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on some embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

The task processing method provided in the present application is applied to a task processing system, which is an edge network system. Referring to FIG1 , FIG1 is a schematic diagram of the structure of a task processing system provided in the present application, the task processing system includes an edge network 100 and various computing nodes 200 deployed in the edge network, wherein the edge network 100 is used to allocate resources for computing tasks to allocate them to appropriate computing nodes 100, so as to realize task processing based on the computing resources in the computing nodes 100.

It is understandable that edge computing is mainly aimed at the massive terminal users in the era of the Internet of Everything. Terminal users and their devices generate highly concurrent computing task requests. Since the capabilities of terminal user devices are extremely limited, by utilizing the computing power resources and network bandwidth resources in the edge computing network, users can reasonably offload computing tasks to the edge, and the computing nodes in the edge network will allocate corresponding computing resources to process user requests and complete task processing.

The present application provides a task processing method in some embodiments.

Referring to FIG. 2, FIG. 2 is a flowchart of a task processing method provided in some embodiments of the present application. The task processing method can be applied to computing nodes in an edge network. Each computing node is provided with one or more heterogeneous resources, including It includes the following S101 to S103.

First of all, it should be noted that the task processing method provided in some embodiments of the present application is applied to each computing node in the edge network, that is, the implementation process of S101 to S103 is executed by the computing node in the edge network. In the edge network, each computing node is distributed and deployed in the geographic space, and its resource provider sources are relatively diverse. In other words, the type of computing node is not unique. For example, it can be an infrastructure service provider such as a telecom operator, a cloud computing service provider, and an individual user connected to the Internet.

Each computing node contains one or more heterogeneous resources, such as CPU, GPU, FPGA and ASIC. These heterogeneous resources have distinct functional characteristics, different characteristics in computing power, power consumption, etc., and can adapt to different types of computing tasks. For example, CPU is good at processing tasks with sequential logic structure, while GPU has strong processing ability for parallel tasks, and FPGA has good processing performance for real-time computing tasks in some special scenarios. Therefore, when facing different computing tasks, the processing capabilities of heterogeneous resources are also different. In one possible implementation, the heterogeneous resources in each computing node in the edge network may include CPU resources, GPU resources, and FPGA resources.

S101: receiving a resource allocation strategy issued by an edge network; the resource allocation strategy includes a processing amount of a target task allocated to each heterogeneous resource;

This step can achieve the acquisition of resource allocation strategy, that is, receiving the resource allocation strategy issued by the edge network. Among them, the resource allocation strategy is a strategy for the target task, which can be constructed by the edge network in combination with the task requirements of the target task and the local resource information of each computing node in the edge network, and is used to allocate reasonable computing resources for the target task to facilitate the processing of the target task, which is the computing task initiated by the user end and needs to be processed.

Among them, the resource allocation strategy is equivalent to the initial allocation strategy, which includes the task volume of the target task allocated to each heterogeneous resource in the edge network. As mentioned above, each computing node contains one or more heterogeneous resources, and the processing volume of the target task allocated to each heterogeneous resource is the task volume of the subtasks of the target task allocated to each heterogeneous resource. The task volume can be the proportion of the subtask to the target task, or the number of subtasks. Obviously, when the task volume is the proportion of subtasks, the sum of the proportions of subtasks of all heterogeneous resources is 1; when the task volume is the number of subtasks, the sum of the number of subtasks of all heterogeneous resources is the total task volume of the target task.

It should be noted that the computing nodes that receive the resource allocation strategy issued by the edge network are not unique. They can be computing nodes selected by the edge network based on a certain screening strategy. Therefore, in the edge network, the computing nodes that receive the resource allocation strategy may be one or more, and may be some computing nodes in the edge network or all computing nodes in the edge network. Of course, when all computing nodes in the edge network are not selected, it means that there are no computing nodes in the current edge network that meet the processing requirements of the target task, and the processing can be delayed for a period of time. On this basis, it can be imagined that the heterogeneous resources mentioned in the resource allocation strategy refer to the heterogeneous resources of each selected computing node.

S102: Input the resource allocation strategy and local resource information into a preset strategy learning model for processing to obtain a resource sharing strategy; the resource sharing strategy includes available resources of various heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

This step can realize the calculation of resource sharing strategy based on the preset strategy learning model. First, in each computing node, a strategy learning model that is suitable for itself is pre-created. The input of the model is the resource allocation strategy issued by the edge network and the resource information of the computing node itself (that is, the local resource information mentioned above). The local resource information mainly refers to the available resources that the computing node itself is expected to provide. The output of the model is the resource sharing strategy. Among them, the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal. The task processing value refers to the computing node in the task processing That is, taking the profit demand of computing nodes into consideration, a preset strategy learning model is constructed to maximize the task processing value of computing nodes. In other words, the resource sharing strategy learned based on the preset strategy learning model can maximize the task processing profit of computing nodes, so as to achieve a win-win situation for task requesters and resource providers.

Furthermore, the resource sharing strategy output by the preset strategy learning model is the final allocation strategy, which includes the available resources of various heterogeneous resources within the current computing node. The available resources here refer to the resources provided for processing the current target task, which can be the ratio of the resources provided for processing the current target task in each heterogeneous resource to the available resources actually provided by itself.

Of course, when there are multiple computing nodes that receive the resource allocation strategy, each computing node that receives the resource allocation strategy will output a resource sharing strategy suitable for itself, and each resource sharing strategy clearly indicates the computing resources provided by the heterogeneous resources within its own computing node for the target task. Therefore, the processing of the target task can be achieved based on the computing resources actually provided by the heterogeneous resources in the selected computing nodes.

S103: Execute the target task using the resource sharing strategy.

This step can realize the execution of the target task, that is, the target task is executed based on the resource sharing strategy. Of course, the execution of the target task here can be to execute the target task by utilizing the available resources of the heterogeneous resources indicated in the resource sharing strategy. It can be understood that since the target task may be jointly executed by various heterogeneous resources in multiple computing nodes, the "sharing" in the resource sharing strategy means that the various heterogeneous resources in the various computing nodes that execute the target task can share information during the execution of the target task, thereby ensuring that the target task can be completed and the accuracy of the target task execution result is guaranteed.

It can be seen that the task processing method provided in some embodiments of the present application pre-deploys a policy learning model in each computing node of the edge network, and determines the resource sharing strategy based on the preset policy learning model, thereby utilizing the resource sharing strategy to implement task processing. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and then the computing node learns the resource allocation strategy based on the resource allocation strategy, and both the resource allocation strategy and the resource sharing strategy take the heterogeneous resources in the computing node into account, further realizing the reasonable allocation of heterogeneous resources within the computing node.

In some embodiments, after the above-mentioned resource sharing strategy is used to execute the target task, the following steps may also be included:

Get the execution result of the target task;

Report the execution results to the edge network.

The task processing method provided in some embodiments of the present application can further implement an execution result feedback function. After the target task is executed based on the resource sharing strategy, the execution result of the target task can also be counted so as to report the execution result to the edge network, thereby realizing the feedback of the execution result. Furthermore, the edge network can also feed back the execution result to the initiator of the target task.

In some embodiments, the task processing method may further include the following steps:

Real-time monitoring of local resource information;

As mentioned above, the resource allocation strategy can be constructed by the edge network in combination with the task requirements of the target task and the local resource information of each computing node in the edge network. Therefore, before constructing the resource allocation strategy, the edge network needs to first Get the local resource information.

Among them, the local resource information is published to the edge network by the corresponding computing node, and each computing node publishes the local resource information to the edge network when the local resource information meets the preset conditions. When the local resource information does not meet its own preset conditions, the local resource information will not be published. Therefore, each computing node can monitor its own local resource information in real time, and publish it to the edge network when the local resource information meets the preset conditions. Among them, the preset conditions are the publishing conditions pre-set by the owner of the computing node according to its actual needs, and are not unique. This application does not limit this.

In some embodiments, when the local resource information meets the preset conditions, publishing the local resource information to the edge network includes the following steps:

Determine available resources based on local resource information;

In some embodiments, the present application provides a type of preset condition, namely, the resource ratio of available resources in the computing node reaches a preset threshold, wherein available resources refer to idle resources in the computing node, and resource ratio refers to the proportion of idle resources to total resources.

In the process of real-time statistics of local resource information, available resources can be further determined based on the local resource information, and the resource ratio of available resources can be counted. When the resource ratio reaches the preset threshold, it means that it can provide computing resources for the computing tasks of other terminals, so the local resource information can be published; conversely, when the resource ratio does not reach the preset threshold, it means that it cannot provide computing resources for the computing tasks of other terminals, so the local resource information is not published. Of course, the value of the preset threshold does not affect the implementation of this technical solution, and it can be set by the owner of the computing node according to actual needs, and this application does not limit this.

In some embodiments, the local resource information may include available resources and resource provision time, and the available resources may include currently available resources which may be heterogeneous resources.

The present application provides a type of local resource information in some embodiments. First, in order to realize task processing, the available resources of the computing node need to meet the actual needs of the target task; secondly, the willingness of the owner of the computing node to share resources sometimes fluctuates over time. For example, when a computing node is relatively idle, the owner of the computing node is more inclined to share its idle resources, serve other users, and obtain certain rewards or benefits. Therefore, the resource provision time of the computing node should also meet the actual needs of the target task. Based on this, local resource information can include its own available resources and resource provision time, among which the available resources include the current available resources of various heterogeneous resources within itself.

The present application provides another task processing method in some embodiments.

Refer to Figure 3, which is a flow chart of another task processing method provided in the present application. The task processing method can be applied to an edge network, and each computing node in the edge network is provided with one or more heterogeneous resources, including the following S201 and S202.

First of all, it should be noted that the task processing method provided in some embodiments of the present application is applied to an edge network, that is, the implementation process of S201 and S202 is executed by the edge network.

S201: Determine a target computing node and a resource allocation strategy according to task information of the target task and local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

This step can determine the target computing node and resource allocation strategy. The target computing node is the computing node selected from the edge network to implement the target task processing, and the resource allocation strategy is the strategy for the target task. It is used to allocate reasonable computing resources to the target task in order to facilitate the processing of the target task. The target task is the computing task initiated by the user end that needs to be processed.

During the implementation process, when the edge network receives the target task to be processed, it can first count the task information of the target task, including but not limited to task type, task target, computing resource demand and other information; then count the local resource information of each computing node in the edge network, and the local resource information may include but not limited to the available resources that the computing node itself is expected to provide, the time limit for providing available resources and other information; finally, combine the task information of the target task and the local resource information of each computing node to screen and determine the target computing node from a large number of computing nodes, and build a resource allocation strategy.

Among them, the resource allocation strategy is equivalent to the initial allocation strategy, which includes the task volume of the target task allocated to each heterogeneous resource in the edge network. It can be understood that each computing node contains one or more heterogeneous resources, wherein the processing volume of the target task allocated to each heterogeneous resource is the task volume of the subtasks of the target task allocated to each heterogeneous resource, and the task volume can be the proportion of the subtask to the target task, or the number of subtasks. Obviously, when the task volume is the proportion of subtasks, the sum of the proportions of subtasks of all heterogeneous resources is 1; when the task volume is the number of subtasks, the sum of the number of subtasks of all heterogeneous resources is the total task volume of the target task.

It should be noted that the number of target computing nodes is not unique. It may be one or more. It may be some computing nodes in the edge network or all computing nodes in the edge network. Of course, if no target computing node is determined, it means that there is no computing node in the current edge network that meets the processing requirements of the target task, and the processing can be delayed for a period of time. On this basis, it can be imagined that the heterogeneous resources mentioned in the resource allocation strategy refer to the heterogeneous resources of each target computing node that has been screened and determined.

S202: Send the resource allocation strategy to each target computing node, so that each target computing node uses a preset strategy learning model to process the resource allocation strategy and local resource information, obtains a resource sharing strategy, and uses the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each heterogeneous resource, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

This step can implement the distribution of resource allocation strategies, that is, the resource allocation strategies are distributed to each target computing node, and each target computing node implements the execution of the target task based on the resource allocation strategy. The implementation process of each target computing node executing the target task is as follows:

First, in each computing node of the edge network, a policy learning model adapted to itself is pre-created. For the target computing node, the input of the model is the resource allocation strategy issued by the edge network and the resource information of the target computing node itself (i.e., the local resource information mentioned above). The local resource information mainly refers to the available resources that the target computing node itself is expected to provide. The output of the model is the resource sharing strategy. Among them, the preset policy learning model is obtained by training with the task processing value of the computing node as the optimization goal. The task processing value refers to the benefit obtained by the computing node when performing task processing, that is, taking the benefit demand of the computing node into consideration, and constructing a preset policy learning model that can maximize the task processing value of the computing node. In other words, the resource sharing strategy learned based on the preset policy learning model can maximize the task processing benefit of the computing node, so as to achieve a win-win situation for the task requester and the resource provider.

Furthermore, the resource sharing strategy output by the preset strategy learning model is the final allocation strategy, which includes the available resources of various heterogeneous resources within the current target computing node. The available resources here refer to the resources provided for processing the current target task, which can be the ratio of the resources provided by each heterogeneous resource for processing the current target task to the available resources actually provided by itself.

Of course, when there are multiple target computing nodes, each target computing node will output a resource sharing strategy suitable for itself, and each resource sharing strategy clearly indicates the heterogeneous resources within its own computing node to provide resources for the target task. The computing resources provided by each target computing node can realize the processing of the target task based on the computing resources actually provided by each heterogeneous resource in each target computing node.

Finally, the target task can be executed based on the resource sharing strategy. Of course, the target task here can be executed by using the available resources of the heterogeneous resources indicated in the resource sharing strategy. It can be understood that since the target task may be jointly executed by the heterogeneous resources in multiple target computing nodes, the "sharing" in the resource sharing strategy means that the heterogeneous resources in the target computing nodes that execute the target task can share information during the execution of the target task, thereby ensuring that the target task can be completed and the accuracy of the target task execution result is guaranteed.

In some embodiments, the above-mentioned determination of the target computing node according to the task information of the target task and the local resource information of each computing node may include the following steps:

The present application provides a method for implementing a screening and determination of target computing nodes in some embodiments. First, in order to realize task processing, the available resources of the target computing node need to meet the actual needs of the target task; second, the willingness of the owner of the computing node to share resources sometimes fluctuates over time. For example, when a computing node is relatively idle, the owner of the computing node is more inclined to share its idle resources, serve other users, and obtain certain rewards or benefits. Therefore, the resource provision time of the target computing node should also meet the actual needs of the target task.

On this basis, in the implementation process, after obtaining the local resource information of each computing node, the resource provision time and available resources of the corresponding computing node can be determined according to the local resource information, and then it can be determined whether the two meet the task requirements indicated in the task information of the target task, and the current computing node is confirmed as the target computing node when the task requirements are met. If any type of information does not meet the task requirements, it will not be confirmed as the target computing node.

Of course, the above selection criteria are only one implementation method provided in some embodiments of the present application, and are not unique. They can be set according to the actual needs of the resource provider and the task requester, and the present application does not limit this.

In some embodiments, before determining the target computing node and the resource allocation strategy based on the task information of the target task and the local resource information of each computing node, the following steps may also be included:

It is understandable that the resource allocation strategy is constructed by the edge network based on the task information of the target task and the local resource information of each computing node. Therefore, before constructing the resource allocation strategy, the edge network needs to obtain the local resource information of each computing node. Among them, the local resource information is published to the edge network by the corresponding computing node. Therefore, when the edge network receives the target task, it can directly capture the local resource information published by each computing node.

In addition, each computing node publishes local resource information to the edge network only when the local resource information meets the preset conditions. If the local resource information does not meet its own preset conditions, the local resource information will not be published. The preset conditions are publishing conditions pre-set by the owner of the computing node according to its actual needs, and are not unique. For example, the idle resource ratio can reach above the preset threshold, and the current time can meet the preset resource sharing time.

Feedback the execution results to the initiator of the target task.

The task processing method provided in some embodiments of the present application can further implement the execution result feedback function, and the execution result is the execution result of the target task. For each target computing node, after completing the target task, it can further report the execution result to the edge network; for the edge network, after receiving the execution result uploaded by the target computing node, it can feedback the final execution result to the initiator of the target task.

The present application provides yet another task processing method in some embodiments.

Referring to FIG. 4 , FIG. 4 is a flowchart of another task processing method provided by the present application. The task processing method may include the following steps:

1. Active discovery strategy for heterogeneous computing resources in distributed edge networks. Each distributed edge computing node actively and timely publishes resource status tags to the edge network based on its own operating status and available resources;

2. User task resource combination allocation strategy based on heterogeneous computing. According to the current user's computing task request information and the computing resource status label obtained in the current edge network, the schedulable status of various types of heterogeneous computing resources is counted to obtain the heterogeneous computing resource combination allocation plan for the computing task.

3. Learning heterogeneous computing node resource sharing strategies. Using multi-agent strategy learning technology, each heterogeneous edge node outputs its own computing resource sharing strategy based on the heterogeneous computing resource combination allocation plan given in the previous step.

The implementation process of the above steps is as follows:

1. Active discovery strategy for heterogeneous computing resources in distributed edge networks.

In an edge computing network, each edge computing node is distributed and deployed in geographic space, and its resource providers come from a variety of sources. These edge computing nodes are usually heterogeneous, and the capabilities and types of computing resources that different edge computing nodes can provide may vary. In addition, the owner's willingness to share computing resources also fluctuates over time. Therefore, according to certain strategies, the heterogeneous computing resources in the edge network can be actively shared and discovered.

(1) Edge node computing resource label setting:

The owner of the edge computing node describes the heterogeneous computing resources it owns in the form of a label, which can be expressed as the following tuple:
Label _i ={Thd _i , Time _i , C _i , G _i , F _i };

Among them, Label _i represents the label used by edge computing node i for active resource discovery, Thd _i represents the threshold at which the edge computing node can publish the label, which can be the percentage of all currently available computing resources to the total amount of computing resources actually owned by the edge computing node, and Time _i represents the time period set by the owner for resource sharing. _Ci , _Gi , and _Fi respectively represent the CPU, GPU, and FPGA computing resources currently available to the edge computing node.

(2) Active resource discovery strategy:

On this edge computing node, the resource publishing module is normally run to monitor the utilization of various heterogeneous resources and the current time in real time. When the publishing conditions are met, the module actively publishes the label Label _i to the edge network to indicate that the computing resources on the edge computing node can be shared within the time range Time _i , and the types and sizes of the sharable resources are _Ci , _Gi, and F _i respectively.

2. User task resource combination allocation strategy based on heterogeneous computing.

When a computing task (target task) is generated in the edge network, the task scheduling module allocates resources for the user's computing task based on the available heterogeneous computing resources to improve the utilization of edge network resources, meet the various constraints of the user's computing task, and improve the user's quality of experience.

(1) Heterogeneous computing resource information collection. The task scheduling module first counts the schedulable status of various heterogeneous computing resources based on the tags actively released by each computing node, and selects the computing nodes that can be used during the time period when the user's computing task is completed. Under this condition, it counts whether the heterogeneous computing resources that can be scheduled can meet the user's computing task. If so, it proceeds to the next step. Otherwise, the computing task cannot be met.

(2) Generation of heterogeneous resource combination allocation strategy. Based on the consumption required by various computing resources to process unit computing tasks and the size of user computing tasks, multiple allocation strategies based on heterogeneous computing resource combination can be obtained. Assume that the resource consumption required by the CPU, GPU and FPGA to process unit computing tasks is real number c, g, f, respectively, the total task amount of a computing task generated at time t is real number m _t , and the computing task can be divided into multiple real number parts, where the task amounts corresponding to the CPU, GPU and FPGA are real numbers x, y, z, respectively. Under this condition, the total consumption of computing task t is:
Cost _t = cx + gy + fz;

The corresponding restrictions are:
T _t ＜T _thd ；
m _t = x + y + z;

in, represents the true value of the constraint item for this computing task, They represent the constraint factors of heterogeneous computing resources such as CPU, GPU, and FPGA corresponding to the computing task at time t. The combination allocation strategy of heterogeneous computing resources should reduce the total consumption Cost _t as much as possible, optimize the average cost of the edge network, and improve the user experience quality under the condition of satisfying the task constraint item T _thd and the total computing task m _t . The process of solving the optimal solution of the objective function can be expressed as a linear programming problem. The optimal heterogeneous computing resource combination strategy can be obtained by solving the above objective function using a linear programming solver.

3. Learning resource sharing strategies for heterogeneous computing nodes.

Using multi-agent strategy learning technology, each edge computing node learns its own resource allocation strategy based on the optimal resource combination strategy obtained in the previous step. In the current set of edge computing nodes, each edge computing node shares part of its computing resources for the processing of the current computing task and obtains corresponding benefits.

For edge computing node i in the set, let its resource sharing strategy for the current computing task be:

in, They are the proportions of various heterogeneous computing resources (CPU, GPU, FPGA) that are expected to be allocated to the current computing task to their own currently available resources. They are all real numbers between 0 and 1 and satisfy:

That is, the computing resources shared by each edge computing node can meet the optimal resource allocation strategy for computing tasks. The strategy learning goal of each edge computing node is to generate the optimal resource sharing strategy a _i so that it can maximize the current benefit r _i :

in, Respectively represent the rewards obtained by sharing their heterogeneous resources with the current computing task. The reward represents the benefits that resource owners can obtain when they provide their own heterogeneous computing resources for task processing, and each resource owner seeks to maximize the benefits they obtain. The values represent the reward that can be obtained when a unit of heterogeneous computing resources processes a task, that is, the basic benefit when sharing a certain type of resource, which is a positive real number. Generally speaking, heterogeneous resources have different processing capabilities for different types of computing tasks. Therefore, it is necessary to further improve the efficiency of resource use and the benefit level of resource owners by setting the unit benefit of heterogeneous resources when facing computing tasks.

Using multi-agent strategy learning technology, the resource sharing strategy of each edge computing node is learned, and a centralized training and distributed execution mechanism is adopted. Set the strategy learning network (preset strategy learning model) of the edge computing node to _Ni , The network parameters are ω _i , and its input is the heterogeneous computing resource combination allocation strategy s ₁ = {x, y, z} _opt of the current computing task and the currently available heterogeneous computing resources s ₂ = {C _i , G _i , F _i } of the edge computing node. The output of the policy learning network is the resource sharing decision

In addition, each edge computing node sets a separate value evaluation network _Qi , whose parameters are _θi . In order to obtain the optimal shared decision output, the parameters of _Ni need to be trained. The training process of each edge computing node is as follows:

(1) The state information s _i = {s ₁ , s ₂ } obtained by edge computing node i is used as the input of the local decision network to obtain the current output a _i , execute action a _i , and the system is converted to the new state s ′ _i and the corresponding benefit _ri , and {s _i , a _i , _ri , s ′ _i } is saved in the set D;

(2) When the number of records in set D is greater than M, sample M times from set D and calculate the target value:

Where s ^′m represents the next environmental information observed by all edge computing nodes, γ is a real number between 0 and 1, representing the discount factor;

(3) Update the parameters θ _i of the value evaluation network _Qi of each edge computing node according to the minimization loss function. The minimization loss function is:

Among them, s ^m represents the current environment information observed by all edge computing nodes;

(4) The update gradient of the strategy learning network _Ni is:

(5) After the training converges or reaches the specified number of training times, the training is completed, and _{Ni is} deployed in the edge computing node to generate the resource sharing strategy of the current edge computing node.

It can be seen that the task processing method provided in some embodiments of the present application pre-deploys a policy learning model in each computing node of the edge network, and determines the resource sharing strategy based on the preset policy learning model, thereby utilizing the resource sharing strategy to implement task processing. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the obtained resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and the computing node learns the resource allocation strategy based on the resource allocation strategy, and Both the resource allocation strategy and resource sharing strategy take the heterogeneous resources in the computing nodes into consideration, further realizing the reasonable allocation of heterogeneous resources within the computing nodes.

The present application provides a task processing system in some embodiments.

As shown in FIG1 , the task processing system may include an edge network 100 and computing nodes 200 deployed in the edge network 100 , each computing node 200 being provided with one or more heterogeneous resources, wherein:

The edge network 100 is used to determine the target computing node 200 and the resource allocation strategy according to the task information of the target task and the local resource information of each computing node 200, and send the resource allocation strategy to each target computing node 200; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The target computing node 200 is used to process the resource allocation strategy and local resource information using a preset strategy learning model, obtain a resource sharing strategy, and use the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of various heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

It can be seen that the task processing system provided in some embodiments of the present application pre-deploys a policy learning model in each computing node of the edge network, and determines the resource sharing strategy based on the preset policy learning model, thereby utilizing the resource sharing strategy to implement task processing. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and then the computing node learns the resource allocation strategy based on the resource allocation strategy, and both the resource allocation strategy and the resource sharing strategy take the heterogeneous resources in the computing node into account, further realizing the reasonable allocation of heterogeneous resources within the computing node.

For an introduction to the system provided in some embodiments of the present application, please refer to the above method embodiments, and the present application will not go into details here.

The present application provides a task processing device in some embodiments.

Referring to FIG. 5 , FIG. 5 is a schematic diagram of the structure of a task processing device provided by the present application. The task processing device can be applied to computing nodes in an edge network. Each computing node is provided with one or more heterogeneous resources, including:

The receiving module 1 is used to receive the resource allocation strategy sent by the edge network; the resource allocation strategy includes the processing capacity of each heterogeneous resource allocated to the target task;

Processing module 2 is used to input the resource allocation strategy and local resource information into the preset strategy learning model for processing to obtain the resource sharing strategy; the resource sharing strategy includes the available resources of each heterogeneous resource, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

The execution module 3 is used to execute the target task by utilizing the resource sharing strategy.

It can be seen that the task processing device provided in some embodiments of the present application pre-deploys a policy learning model in each computing node of the edge network, and determines the resource sharing strategy based on the preset policy learning model, thereby utilizing the resource sharing strategy to implement task processing. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the obtained resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and then the computing node learns the resource allocation strategy based on the resource allocation strategy, and both the resource allocation strategy and the resource sharing strategy take the heterogeneous resources in the computing node into account, further realizing the computing node. Reasonable allocation of internal heterogeneous resources.

In some embodiments, the task processing device may further include a reporting module, which is used to obtain the execution result of the target task after executing the target task using the resource sharing strategy described above; and report the execution result to the edge network.

In some embodiments, the above heterogeneous resources may include CPU resources, GPU resources, and FPGA resources.

In some embodiments, the task processing device may further include a publishing module for real-time monitoring of local resource information; when the local resource information meets a preset condition, the local resource information is published to the edge network.

In some embodiments, the publishing module may be used to determine available resources based on local resource information; when the resource ratio of available resources reaches a preset threshold, the local resource information is published to the edge network.

In some embodiments, the local resource information may include available resources and resource provision time, and the available resources may include currently available resources of various heterogeneous resources.

For an introduction to the apparatus provided in some embodiments of the present application, please refer to the above method embodiments, and the present application will not elaborate on them here.

The present application provides another task processing device in some embodiments.

Referring to FIG. 6 , FIG. 6 is a schematic diagram of the structure of another task processing device provided by the present application, which can be applied to an edge network, and each computing node in the edge network is provided with one or more heterogeneous resources, including:

Determining module 4, used to determine the target computing node and resource allocation strategy according to the task information of the target task and the local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The sending module 5 is used to send the resource allocation strategy to each target computing node, so that each target computing node uses the preset strategy learning model to process the resource allocation strategy and local resource information, obtains the resource sharing strategy, and uses the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each heterogeneous resource, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.

It can be seen that the task processing device provided in some embodiments of the present application pre-deploys a policy learning model in each computing node of the edge network, and determines the resource sharing strategy based on the preset policy learning model, thereby utilizing the resource sharing strategy to implement task processing. Since the preset policy learning model is trained with the task processing value of the computing node as the optimization goal, task processing based on the resource sharing strategy can maximize the task processing value of the computing node and ensure that the task is processed, thereby achieving a win-win situation for the task requester and the resource provider. In addition, the resource allocation strategy is created by the edge network, and then the computing node learns the resource allocation strategy based on the resource allocation strategy, and both the resource allocation strategy and the resource sharing strategy take the heterogeneous resources in the computing node into account, further realizing the reasonable allocation of heterogeneous resources within the computing node.

In some embodiments, the determination module 4 may be used to determine resource provision time and available resources based on local resource information; when both resource provision time and available resources meet the task requirements indicated by the task information, the computing node is determined to be the target computing node.

In some embodiments, the task processing device may also include a statistical module for counting the local resource information published by each computing node before determining the target computing node and resource allocation strategy based on the task information of the target task and the local resource information of each computing node; wherein the local resource information is published to the edge network by the computing node when the local resource information meets preset conditions.

In some embodiments, the task processing device may further include a feedback module for obtaining the target computing node uploaded The execution result of the target task; the execution result is fed back to the initiator of the target task.

The present application provides a task processing device in some embodiments.

Referring to FIG. 7 , FIG. 7 is a schematic diagram of the structure of a task processing device provided by the present application, and the task processing device may include:

Memory for storing computer programs;

The processor can implement the steps of any one of the above-mentioned task processing methods when being used to execute a computer program.

As shown in Fig. 7, it is a schematic diagram of the composition structure of the task processing device, which may include: a processor 10, a memory 11, a communication interface 12 and a communication bus 13. The processor 10, the memory 11 and the communication interface 12 all communicate with each other through the communication bus 13.

In some embodiments of the present application, the processor 10 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic devices.

The processor 10 may call a program stored in the memory 11 , and the processor 10 may execute operations in the embodiment of the task processing method.

The memory 11 is used to store one or more programs, which may include program codes, and the program codes include computer operation instructions. In some embodiments of the present application, the memory 11 stores at least a program for implementing the following functions:

Use resource sharing strategies to execute target tasks;

or,

In a possible implementation, the memory 11 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function, etc.; the data storage area may store data created during use.

In addition, the memory 11 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

The communication interface 12 may be an interface of a communication module, and is used to connect to other devices or systems.

Of course, it should be noted that the structure shown in FIG. 7 does not constitute a task processing device in some embodiments of the present application. In actual applications, the task processing device may include more or fewer components than those shown in FIG. 7 , or may combine certain components.

The present application provides a non-volatile computer-readable storage medium in some embodiments.

As shown in FIG8 , in some embodiments of the present application, a non-volatile computer-readable storage medium 80 stores a computer program 810 , and when the computer program 810 is executed by a processor, the steps of any of the above-mentioned task processing methods can be implemented.

The non-volatile computer-readable storage medium may include: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

For an introduction to the non-volatile computer-readable storage medium provided in some embodiments of the present application, please refer to the above method embodiments, and the present application will not elaborate on them here.

The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

The technical solution provided by the present application is described in detail above. The principle and implementation method of the present application are described in detail using specific examples herein, and the description of the above embodiments is only used to help understand the method and core idea of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the present application.

Claims

A task processing method, characterized in that it is applied to computing nodes in an edge network, each of the computing nodes is provided with one or more heterogeneous resources, and the method comprises:

Receiving a resource allocation strategy issued by the edge network; the resource allocation strategy includes a processing amount of a target task to which each of the heterogeneous resources is allocated;

The resource allocation strategy and local resource information are input into a preset strategy learning model for processing to obtain a resource sharing strategy; the resource sharing strategy includes the available resources of each of the heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

The target task is executed using the resource sharing strategy.
The method according to claim 1, characterized in that after executing the target task using the resource sharing strategy, it also includes:

Obtaining the execution result of the target task;

The execution result is reported to the edge network.
The method according to claim 1 is characterized in that the heterogeneous resources include central processing unit (CPU) resources, graphics processing unit (GPU) resources, and field programmable gate array (FPGA) resources.
The method according to claim 3 is characterized in that the heterogeneous resources also include application-specific integrated circuit ASIC resources.
The method according to any one of claims 1 to 4, further comprising:

Performing real-time monitoring on the local resource information;

When the local resource information meets a preset condition, the local resource information is published to the edge network.
The method according to claim 5, characterized in that when the local resource information meets a preset condition, publishing the local resource information to the edge network comprises:

Determining available resources according to the local resource information;

When the resource proportion of the available resources reaches a preset threshold, the local resource information is published to the edge network.
The method according to claim 5 is characterized in that the local resource information includes available resources and resource provision time, and the available resources include currently available resources of each of the heterogeneous resources.
The method according to claim 6, characterized in that

The available resources refer to idle resources in the computing nodes, and the resource proportion refers to the ratio of idle resources to total resources.
The method according to claim 1 is characterized in that the task volume is the target task assigned to each The workload of subtasks of heterogeneous resources.
The method according to claim 9 is characterized in that the task amount is the proportion of the subtask to the target task, or the number of the subtasks.
A task processing method, characterized in that it is applied to an edge network, each computing node in the edge network is provided with one or more heterogeneous resources, and the method comprises:

Determine the target computing node and the resource allocation strategy according to the task information of the target task and the local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The resource allocation strategy is sent to each of the target computing nodes, so that each of the target computing nodes uses a preset strategy learning model to process the resource allocation strategy and the local resource information, obtains a resource sharing strategy, and uses the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each of the heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.
The method according to claim 11, characterized in that the step of determining the target computing node according to the task information of the target task and the local resource information of each computing node comprises:

Determining resource provision time and available resources according to the local resource information;

When the resource provision time and the available resources both meet the task requirements indicated by the task information, the computing node is determined to be the target computing node.
The method according to claim 11, characterized in that before determining the target computing node and the resource allocation strategy based on the task information of the target task and the local resource information of each computing node, it also includes:

The local resource information published by each of the computing nodes is counted; wherein the local resource information is published by the computing node to the edge network when the local resource information meets a preset condition.
The method according to claim 11, further comprising:

Obtaining the execution result of the target task uploaded by the target computing node;

The execution result is fed back to the initiator of the target task.
A task processing system, characterized in that it comprises an edge network and computing nodes deployed in the edge network, each computing node is provided with one or more heterogeneous resources, wherein:

The edge network is used to determine the target computing node and the resource allocation strategy according to the task information of the target task and the local resource information of each computing node, and send the resource allocation strategy to each target computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

The target computing node is used to use a preset strategy learning model to learn the resource allocation strategy and the local resource The source information is processed to obtain a resource sharing strategy, and the target task is executed using the resource sharing strategy; the resource sharing strategy includes the available resources of each of the heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.
The system according to claim 15, characterized in that

The edge network is also used to feed back the execution result to the initiator of the target task.
A task processing device, characterized in that it is applied to computing nodes in an edge network, each of the computing nodes is provided with one or more heterogeneous resources, and the device comprises:

A receiving module, configured to receive a resource allocation strategy issued by the edge network; the resource allocation strategy includes a processing amount of a target task to which each of the heterogeneous resources is allocated;

A processing module, used for inputting the resource allocation strategy and local resource information into a preset strategy learning model for processing to obtain a resource sharing strategy; the resource sharing strategy includes the available resources of each of the heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization target;

An execution module is used to execute the target task using the resource sharing strategy.
A task processing device, characterized in that it is applied to an edge network, each computing node in the edge network is provided with one or more heterogeneous resources, and the device comprises:

A determination module, configured to determine a target computing node and a resource allocation strategy according to the task information of the target task and the local resource information of each computing node; the resource allocation strategy includes the processing amount of each heterogeneous resource allocated to the target task;

A sending module is used to send the resource allocation strategy to each of the target computing nodes, so that each of the target computing nodes uses a preset strategy learning model to process the resource allocation strategy and the local resource information, obtain a resource sharing strategy, and use the resource sharing strategy to execute the target task; the resource sharing strategy includes the available resources of each of the heterogeneous resources, and the preset strategy learning model is trained with the task processing value of the computing node as the optimization goal.
A task processing device, comprising:

Memory for storing computer programs;

A processor, configured to implement the steps of the task processing method according to any one of claims 1 to 14 when executing the computer program.
A non-volatile computer-readable storage medium, characterized in that a computer program is stored on the non-volatile computer-readable storage medium, and when the computer program is executed by a processor, the steps of the task processing method according to any one of claims 1 to 14 are implemented.