AI-based computing resource collaboration implementation method and system

Abstract

This invention relates to the field of data processing technology and discloses an AI-based method and system for collaborative computing resources. By receiving computing resource sharing requests submitted by suppliers, and after verifying the suppliers and their requests, the computing resources available to the suppliers are added to a computing resource pool. This allows for the fusion of available computing resources. Furthermore, by combining the computing resource demand data of different demanders, the computing resources in the pool are organized to form combinations that meet diverse needs, improving the accuracy of pre-allocation of computing resources. Once a demander receives a collaborative request and is verified, a suitable combination of computing resources is matched based on their demand data, and then computing resources are provided to them. This improves the efficiency and accuracy of providing suitable computing resources to demanders and also helps to increase resource utilization.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for collaborative implementation of computing resources based on AI. Background Technology

[0002] With the rapid development of large-scale artificial intelligence models, the demand for network technology in intelligent computing centers is also increasing, such as higher bandwidth requirements and lower latency requirements.

[0003] However, in existing technologies, the allocation of computing resources in intelligent computing center networks mainly adopts a static resource allocation strategy, specifically: providing a fixed amount of computing resources to users within a local computing resource pool. This singular, localized allocation method easily leads to inappropriate computing resources being provided to users, and the efficiency and accuracy of resource provision are low, which is detrimental to improving resource utilization. Therefore, proposing a technical solution to improve the efficiency and accuracy of providing appropriate computing resources to users is particularly important. Summary of the Invention

[0004] This invention provides an AI-based method and system for collaborative computing resources, which can improve the efficiency and accuracy of providing suitable computing resources to users and improve resource utilization.

[0005] The first aspect of this invention discloses a method for collaborative implementation of computing resources based on AI, the method comprising:

[0006] Receive resource sharing request data submitted by a supply object with computing power resources. The resource sharing request data is used to request the provision of computing power resources to a pre-built computing power resource pool for scheduling by the computing power resource pool.

[0007] After the supply object and the resource sharing request data submitted by the supply object are verified, the computing power resources that the supply object can provide are added to the computing power resource pool.

[0008] Based on the computing power resource demand data corresponding to different demand objects, the computing power resources in the computing power resource pool are sorted out to determine the computing power resource combination that can meet the different computing power resource demand data.

[0009] When a resource collaboration request data of a certain requested object is received, the requested object is authenticated;

[0010] After the object of the demand is authenticated, a target computing resource combination that matches the object of the demand is determined from all the computing resource combinations based on the computing resource demand data corresponding to the object of the demand; and computing resources are provided to the object of the demand based on the target computing resource combination.

[0011] As an optional implementation, in the first aspect of the present invention, the step of sorting out the computing resources in the computing resource pool based on the obtained computing resource demand data corresponding to different demand objects, in order to determine the computing resource combination that can meet the different computing resource demand data, includes:

[0012] Based on the obtained computing power resource demand data corresponding to different demand objects, analyze the performance demand data of each demand object in terms of performance, and based on the computing power resource demand data corresponding to each demand object, analyze the energy efficiency demand data of each demand object in terms of energy efficiency.

[0013] Based on the cloud-edge-device collaborative architecture, according to the performance requirement data corresponding to each of the aforementioned demand objects, corresponding nodes are associated in the computing power resource pool to obtain the associated node data of each of the aforementioned demand objects in the computing power resource pool. The associated node data includes edge node data or cloud center data.

[0014] Based on the energy efficiency requirement data corresponding to each of the aforementioned requirement objects, each of the aforementioned requirement objects is matched with the computing power resource areas of different energy efficiency levels obtained by dividing the computing power resource pool according to energy efficiency, so as to obtain the target computing power resource area that matches each of the aforementioned requirement objects.

[0015] Based on the associated node data of each demand object in the computing power resource pool, and the target computing power resource region that matches the demand object, a combination of computing power resources that can meet the computing power resource demand data corresponding to the demand object is determined from the computing power resource pool.

[0016] As an optional implementation, in a first aspect of the present invention, determining a combination of computing resources from the computing resource pool that can satisfy the computing resource demand data corresponding to each demand object, based on the associated node data of each demand object in the computing resource pool and the target computing resource region matching the demand object, includes:

[0017] Based on the computing power resource demand data corresponding to each demand object, analyze the resource allocation priority data corresponding to each demand object, and generate priority queue data based on all demand objects and the resource allocation priority data corresponding to all demand objects.

[0018] Based on the priority queue data, the target demand objects of the computing resources to be sorted out are determined;

[0019] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, a computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, and the operation of determining the target demand object of the current computing power resource to be sorted based on the priority queue data is re-executed until the required computing power resource combination is sorted for all demand objects in the priority queue data.

[0020] As an optional implementation, in a first aspect of the present invention, determining a combination of computing resources from the computing resource pool that can satisfy the computing resource demand data corresponding to the target demand object, based on the associated node data of the target demand object in the computing resource pool and the target computing resource region matching the target demand object, includes:

[0021] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource region, and each computing power resource has corresponding computing power resource data;

[0022] Based on the computing power resource data corresponding to each of the candidate computing power resources, analyze the scope of computing power service targets for each of the candidate computing power resources;

[0023] Based on the scope of computing power service targets of all the candidate computing power resources, determine the combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand object from all the candidate computing power resources.

[0024] As an optional implementation, in a first aspect of the present invention, determining all candidate computing resources matching the target demand object from all computing resources in the target computing resource region based on the associated node data of the target demand object in the computing resource pool and the target computing resource region matching the target demand object includes:

[0025] Based on the computing power resource demand data corresponding to the target demand object, determine the expected computing power deployment range corresponding to the target demand object;

[0026] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all computing power resources under the desired computing power deployment range are obtained from all computing power resources in the target computing power resource region, as all alternative computing power resources that match the target demand object.

[0027] As an optional implementation, in a first aspect of the present invention, the performance requirement data includes a first sub-requirement data for computing resource utilization and / or a second sub-requirement data for computing-network convergence scheduling delay.

[0028] Furthermore, the cloud-edge-device collaborative architecture, based on the performance requirement data corresponding to each of the aforementioned requirement objects, associates corresponding nodes in the computing resource pool to obtain the associated node data of each of the aforementioned requirement objects in the computing resource pool, including:

[0029] Based on the computing power resource demand data corresponding to each demand object, determine the first demand level for the first sub-demand data and the second demand level for the second sub-demand data respectively.

[0030] Based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data.

[0031] For each of the aforementioned demand objects, the associated node data of the demand object in the computing power resource pool is determined based on the first node data corresponding to the demand object, the first demand level for the first node data, the second node data corresponding to the demand object, and the second demand level for the second node data.

[0032] As an optional implementation, in a first aspect of the present invention, determining a target computing resource combination matching the demand object from all said computing resource combinations based on the computing resource demand data corresponding to the demand object includes:

[0033] From all the aforementioned computing resource combinations, determine the computing resource combination that can meet the computing resource requirement data corresponding to the object with the requirement, and use it as the target computing resource combination matching the object with the requirement; or...

[0034] Based on the resource collaboration request data of the demand object, predict the resource demand forecast data corresponding to the demand object;

[0035] Based on the resource demand forecast data corresponding to the demand object and the computing power resource demand data for each of the aforementioned demand objects, calculate the demand matching degree of the demand object with respect to each of the aforementioned computing power resource demand data.

[0036] Based on the matching degree of the demand for all the computing power resource demand data, target computing power resource demand data with a matching degree greater than or equal to a preset matching degree are determined from all the computing power resource demand data; and the computing power resource combination that can meet the target computing power resource demand data is determined as the target computing power resource combination that matches the demand object.

[0037] A second aspect of this invention discloses an AI-based computing resource collaborative implementation system, the system comprising:

[0038] The data receiving module is used to receive resource sharing request data submitted by a supply object with computing power resources. The resource sharing request data is used to request computing power resources to be provided to a pre-built computing power resource pool for scheduling by the computing power resource pool.

[0039] The resource sharing module is used to add the computing power resources that the supply object can provide to the computing power resource pool after the verification of the supply object and the resource sharing request data submitted by the supply object is passed.

[0040] The resource sorting module is used to sort out the computing resources in the computing resource pool based on the computing resource demand data corresponding to different demand objects, so as to determine the computing resource combination that can meet the computing resource demand data of different computing resource objects.

[0041] The data authentication module is used to authenticate the requesting object when it receives resource collaboration request data from a certain requesting object.

[0042] The resource matching module is used to determine the target computing resource combination that matches the demand object from all the computing resource combinations after the demand object has been authenticated, based on the computing resource demand data corresponding to the demand object.

[0043] The resource scheduling module is used to provide computing resources to the target object based on the target computing resource combination.

[0044] As an optional implementation, in a second aspect of the present invention, the resource sorting module sorts the computing resources in the computing resource pool according to the computing resource demand data corresponding to different demand objects, so as to determine the computing resource combination that can meet the different computing resource demand data. Specifically, this includes:

[0045] Based on the obtained computing power resource demand data corresponding to different demand objects, analyze the performance demand data of each demand object in terms of performance, and based on the computing power resource demand data corresponding to each demand object, analyze the energy efficiency demand data of each demand object in terms of energy efficiency.

[0046] Based on the cloud-edge-device collaborative architecture, according to the performance requirement data corresponding to each of the aforementioned demand objects, corresponding nodes are associated in the computing power resource pool to obtain the associated node data of each of the aforementioned demand objects in the computing power resource pool. The associated node data includes edge node data or cloud center data.

[0047] Based on the energy efficiency requirement data corresponding to each of the aforementioned requirement objects, each of the aforementioned requirement objects is matched with the computing power resource areas of different energy efficiency levels obtained by dividing the computing power resource pool according to energy efficiency, so as to obtain the target computing power resource area that matches each of the aforementioned requirement objects.

[0048] Based on the associated node data of each demand object in the computing power resource pool, and the target computing power resource region that matches the demand object, a combination of computing power resources that can meet the computing power resource demand data corresponding to the demand object is determined from the computing power resource pool.

[0049] As an optional implementation, in a second aspect of the present invention, the resource sorting module determines, based on the associated node data of each demand object in the computing power resource pool and the target computing power resource region matching the demand object, a combination of computing power resources that can satisfy the computing power resource demand data corresponding to the demand object from the computing power resource pool. Specifically, this includes:

[0050] Based on the computing power resource demand data corresponding to each demand object, analyze the resource allocation priority data corresponding to each demand object, and generate priority queue data based on all demand objects and the resource allocation priority data corresponding to all demand objects.

[0051] Based on the priority queue data, the target demand objects of the computing resources to be sorted out are determined;

[0052] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, a computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, and the operation of determining the target demand object of the current computing power resource to be sorted based on the priority queue data is re-executed until the required computing power resource combination is sorted for all demand objects in the priority queue data.

[0053] As an optional implementation, in a second aspect of the present invention, the resource sorting module determines, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource region matching the target demand object, a combination of computing power resources that can satisfy the computing power resource demand data corresponding to the target demand object from the computing power resource pool. Specifically, this includes:

[0054] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource region, and each computing power resource has corresponding computing power resource data;

[0055] Based on the computing power resource data corresponding to each of the candidate computing power resources, analyze the scope of computing power service targets for each of the candidate computing power resources;

[0056] Based on the scope of computing power service targets of all the candidate computing power resources, determine the combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand object from all the candidate computing power resources.

[0057] As an optional implementation, in a second aspect of the present invention, the resource sorting module determines all candidate computing resources matching the target demand object from all computing resources in the target computing resource region based on the associated node data of the target demand object in the computing resource pool and the target computing resource region that matches the target demand object. Specifically, this includes:

[0058] Based on the computing power resource demand data corresponding to the target demand object, determine the expected computing power deployment range corresponding to the target demand object;

[0059] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all computing power resources under the desired computing power deployment range are obtained from all computing power resources in the target computing power resource region, as all alternative computing power resources that match the target demand object.

[0060] As an optional implementation, in a second aspect of the invention, the performance requirement data includes a first sub-requirement data for computing resource utilization and / or a second sub-requirement data for computing-network convergence scheduling delay.

[0061] Furthermore, the resource management module, based on a cloud-edge-device collaborative architecture, associates corresponding nodes in the computing resource pool according to the performance requirement data corresponding to each of the aforementioned requirement objects. The specific methods for obtaining the associated node data of each requirement object in the computing resource pool include:

[0062] Based on the computing power resource demand data corresponding to each demand object, determine the first demand level for the first sub-demand data and the second demand level for the second sub-demand data respectively.

[0063] Based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data.

[0064] For each of the aforementioned demand objects, the associated node data of the demand object in the computing power resource pool is determined based on the first node data corresponding to the demand object, the first demand level for the first node data, the second node data corresponding to the demand object, and the second demand level for the second node data.

[0065] As an optional implementation, in a second aspect of the present invention, the method by which the resource matching module determines the target computing power resource combination that matches the demand object from all the computing power resource combinations based on the computing power resource demand data corresponding to the demand object specifically includes:

[0066] From all the aforementioned computing resource combinations, determine the computing resource combination that can meet the computing resource requirement data corresponding to the object with the requirement, and use it as the target computing resource combination matching the object with the requirement; or...

[0067] Based on the resource collaboration request data of the demand object, predict the resource demand forecast data corresponding to the demand object;

[0068] Based on the resource demand forecast data corresponding to the demand object and the computing power resource demand data for each of the aforementioned demand objects, calculate the demand matching degree of the demand object with respect to each of the aforementioned computing power resource demand data.

[0069] Based on the matching degree of the demand for all the computing power resource demand data, target computing power resource demand data with a matching degree greater than or equal to a preset matching degree are determined from all the computing power resource demand data; and the computing power resource combination that can meet the target computing power resource demand data is determined as the target computing power resource combination that matches the demand object.

[0070] A third aspect of this invention discloses another AI-based computing resource collaborative implementation system, the system comprising:

[0071] Memory containing executable program code;

[0072] A processor coupled to the memory;

[0073] The processor calls the executable program code stored in the memory to execute some or all of the steps in the AI-based computing resource collaborative implementation method described in any of the first aspects of the present invention.

[0074] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the AI-based computing resource collaborative implementation method described in any of the first aspects of the present invention.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] In this embodiment of the invention, by receiving resource sharing request data submitted by a supplier with computing power resources, and after verifying the supplier and its submitted resource sharing request data, the computing power resources that the supplier can provide are added to the computing power resource pool. This allows for the fusion of available computing power resources, which is beneficial to improving the accuracy and reliability of computing power resource sharing. Furthermore, by combining the computing power resource demand data of different demanders, the computing power resources in the computing power resource pool are sorted to form computing power resource combinations that can meet the different computing power resource demand data. This improves the accuracy of pre-allocation of computing power resources and is beneficial to improving the matching accuracy between demanders and computing power resource combinations. After actually receiving resource collaboration request data from a demander and authenticating it, a suitable computing power resource combination is matched for the demander based on the corresponding computing power resource demand data, and then computing power resources are provided to it. This method of pre-allocating computing power resource combinations can improve the efficiency, speed, and timeliness of actual computing power resource allocation. This not only improves the efficiency and accuracy of providing suitable computing power resources to demanders but also helps to improve resource utilization. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a flowchart illustrating a method for collaborative implementation of AI-based computing resources disclosed in an embodiment of the present invention.

[0079] Figure 2 This is a flowchart illustrating another AI-based collaborative computing resource implementation method disclosed in an embodiment of the present invention.

[0080] Figure 3This is a schematic diagram of the structure of an AI-based computing resource collaborative realization system disclosed in an embodiment of the present invention;

[0081] Figure 4 This is a schematic diagram of another AI-based computing resource collaborative implementation system disclosed in an embodiment of the present invention. Detailed Implementation

[0082] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0084] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0085] This invention discloses an AI-based method and system for collaborative computing resources. It receives resource sharing requests from providers of computing resources and, after verifying the providers and their requests, adds the available computing resources to a computing resource pool. This integration of available resources improves the accuracy and reliability of resource sharing. Furthermore, by combining the computing resource needs of different users, the system organizes the resources in the pool to create combinations that meet diverse needs, improving the accuracy of pre-allocation and matching. Upon receiving and verifying a resource collaboration request from a user, the system matches a suitable computing resource combination based on the user's needs and provides the resources. This pre-allocation improves the efficiency, speed, and timeliness of actual computing resource allocation, enhancing both the efficiency and accuracy of providing appropriate resources and improving resource utilization. Detailed explanations follow.

[0086] Example 1

[0087] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for collaborative implementation of AI-based computing resources disclosed in an embodiment of the present invention. Wherein, Figure 1 The described AI-based computing resource collaboration method can be applied to an AI-based computing resource collaboration system. This system may include collaborative devices or collaborative servers, where the collaborative server may be a cloud server or a local server; this embodiment of the invention does not impose limitations. Figure 1 As shown, this AI-based method for collaborative computing resources can include the following operations:

[0088] 101. Receive resource sharing request data submitted by suppliers with computing power resources.

[0089] In this embodiment of the invention, the resource sharing request data is used to request computing resources to be provided to a pre-built computing resource pool for scheduling by the computing resource pool.

[0090] Optionally, the resource sharing request data may include one or more combinations of hardware data, software data, and terms of service data. Hardware data may be at least one of the following: CPU / GPU model, quantity, memory size, storage space, network bandwidth, etc.; software data may be data such as a high-performance network framework (used to support large-scale distributed computing to meet the high bandwidth, low latency, and high reliability requirements of intelligent computing centers), pre-installed operating systems, etc.; terms of service data may be at least one of the following: the time period for which services can be provided, pricing method, service level agreement (SLA) commitments, security policies, etc., and this embodiment of the invention does not impose limitations.

[0091] Optionally, the supply target can be one or more, and this embodiment of the invention does not limit the number of targets.

[0092] 102. After verifying the data of the supply object and the resource sharing request submitted by the supply object, add the computing power resources that the supply object can provide to the computing power resource pool.

[0093] In this embodiment of the invention, specifically, based on the resource sharing request data submitted by the supplier, the identity data of the supplier is determined. The identity data may be the supplier's identity identifier, creditworthiness, etc. Based on the supplier's identity data, the supplier is authenticated (e.g., verifying the supplier's identity is legitimate and trustworthy, and not a malicious node). Furthermore, based on the resource sharing request data submitted by the supplier, the security verification of the resource sharing request data submitted by the supplier is performed (e.g., verifying that the submitted computing resources are genuine, usable, and meet performance standards). If the supplier verification fails, and / or the resource sharing request data submitted by the supplier fails verification, this process can be terminated. If the supplier and the resource sharing request data submitted by the supplier pass verification, the computing resources that the supplier can provide are added to the computing resource pool to realize the sharing of the computing resources that the supplier can provide with other objects.

[0094] 103. Based on the computing power resource demand data corresponding to different demand objects, sort out the computing power resources in the computing power resource pool to determine the computing power resource combination that can meet the different computing power resource demand data.

[0095] In this embodiment of the invention, computing power resource demand data is used to represent the computing power resource demand data of the corresponding demand object. Specifically, based on the computing power resource demand data corresponding to each demand object, the expected computing power deployment range corresponding to each demand object is determined (representing the deployment range of the required computing power resources expected by each demand object), and all computing power resources that can meet the computing power resource demand data corresponding to the demand object are selected from all computing power resources under the expected computing power deployment range corresponding to the demand object, thereby integrating them to obtain a computing power resource combination; or, based on the computing power resource demand data corresponding to each demand object, the relevant demand data of the demand object for different levels are analyzed, such as performance demand data for the performance level and / or energy efficiency demand data for the energy efficiency level, and based on the relevant demand data of the demand object for different levels, all candidate computing power resources matching the demand object are determined from the computing power resource pool, thereby determining all computing power resources within the expected computing power deployment range from all candidate computing power resources, as a computing power resource combination that can meet different computing power resource demand data. This allows for flexible selection of computing resource combinations based on the desired computing power deployment scope and / or the relevant demand data at different levels, which helps improve the flexibility and accuracy of computing resource allocation.

[0096] For example, suppose there are computing resource combinations a, b, and c for computing resource requirements of object A, and object C, where object A is for large language model training, object B is for real-time video rendering, and object C is for cold data storage. In this case, computing resource combination a can consist of computing resources such as a 4x_NVIDIA_A100_80G GPU, a high-speed RDMA network, and a parallel file system; computing resource combination b can consist of computing resources such as a high-performance CPU, a high-end GPU, and large memory; and computing resource combination c can consist of computing resources such as large-capacity hard disk storage and high-throughput bandwidth.

[0097] 104. When a resource collaboration request data is received from a certain requesting object, the requesting object is authenticated.

[0098] In this embodiment of the invention, optionally, the demand object can be the aforementioned supply object, or it can be other objects. Optionally, the resource collaboration request data can include at least one of the following: computing demand data (e.g., what type of CPU / GPU is needed, how much is needed, and how long it will take), task type demand data (e.g., AI training, scientific computing, graphics rendering, etc.), performance demand data (e.g., performance requirements such as latency, bandwidth, and stability), and budget and constraint demand data (e.g., cost budget, time requirements, etc.). This embodiment of the invention does not limit the scope of the data.

[0099] Specifically, based on the resource collaboration request data of the demand object, it is determined whether the demand object belongs to the supply object. If the demand object is found to be a supply object, the demand object is determined to be authenticated. If the demand object is not found to be a supply object, the demand object is directly determined to be unauthenticated. Alternatively, based on the resource collaboration request data of the demand object, the identity data of the demand object and the corresponding qualification verification data (e.g., payment ability verification) are analyzed. Based on the identity data of the demand object, the identity of the demand object is authenticated, and based on the corresponding qualification verification data, the qualification of the resource collaboration request data is authenticated. If the identity authentication of the demand object is successful and the qualification authentication of the resource collaboration request data is successful, the demand object is determined to be authenticated. If the identity authentication of the demand object fails, and / or the qualification authentication of the resource collaboration request data fails, the demand object is determined to be unauthenticated.

[0100] 105. After the object of the demand is certified, the target computing resource combination that matches the object of the demand is determined from all computing resource combinations based on the computing resource demand data corresponding to the object of the demand.

[0101] 106. Provide computing resources to the target user based on the target computing resource combination.

[0102] For example, in a scenario where a SOC system schedules CPU computing resources, the system receives resource sharing request data from at least one CPU with computing resources. After verifying the CPU and its resource sharing request data, the computing resources that the CPU can provide are added to the computing resource pool. Based on the obtained computing resource demand data corresponding to different SOC systems, the computing resources in the pool are sorted to determine the computing resource combinations that can meet the computing resource demands of different SOC systems. When a resource coordination request data from a particular SOC system is received, the SOC system is authenticated. After the SOC system is authenticated, based on its corresponding computing resource demand data, a target computing resource combination matching the SOC system is determined from all computing resource combinations. Computing resources are then provided to the SOC system based on the target computing resource combination. Alternatively,

[0103] In scenarios involving server computing resource scheduling for game testing tasks, the system receives resource sharing request data submitted by servers with computing resources. This request data is used to apply for computing resources to be provided to a pre-built computing resource pool for scheduling. After verifying the server and its resource sharing request data, the computing resources that the server can provide are added to the computing resource pool. Based on the computing resource requirement data corresponding to different game testing tasks, the computing resources in the pool are sorted to determine the computing resource combinations that can meet the different computing resource requirements. When a resource collaboration request data for a specific game testing task is received, the game testing task is authenticated. After the game testing task is authenticated, a target computing resource combination matching the game testing task is determined from all computing resource combinations based on the computing resource requirement data corresponding to the game testing task. Computing resources are then provided to the game testing task based on the target computing resource combination.

[0104] It is evident that implementation Figure 1 The described AI-based computing resource collaboration method receives resource sharing request data submitted by suppliers with computing resources. After verifying the supplier and its resource sharing request data, the method adds the computing resources that the supplier can provide to a computing resource pool. This allows for the fusion of available computing resources, improving the accuracy and reliability of computing resource sharing. Furthermore, by combining the computing resource demand data of different demanders, the method organizes the computing resources in the pool to form computing resource combinations that meet the diverse computing resource needs. This improves the accuracy of pre-allocation of computing resources and enhances the matching accuracy between demanders and computing resource combinations. Upon receiving and authenticating a resource collaboration request data from a specific demander, the method matches a suitable computing resource combination based on the demander's corresponding computing resource demand data and then provides the computing resources. This pre-allocation of computing resource combinations improves the efficiency, speed, and timeliness of actual computing resource allocation. This not only improves the efficiency and accuracy of providing suitable computing resources to demanders but also enhances resource utilization.

[0105] In an optional embodiment, step 105 above, which involves determining the target computing resource combination that matches the demand object from all computing resource combinations based on the computing resource demand data corresponding to the demand object, includes:

[0106] From all available computing resource combinations, identify the computing resource combination that can meet the computing resource requirements of the target user, and use this combination as the target computing resource combination that matches the target user's requirements; or...

[0107] Based on the resource collaboration request data of the demand object, predict the resource demand forecast data corresponding to the demand object;

[0108] Based on the resource demand forecast data corresponding to the demand object and the computing power resource demand data for each demand object, calculate the demand matching degree of the demand object with respect to the computing power resource demand data for each demand object.

[0109] Based on the matching degree of the demand for all computing resource demand data, target computing resource demand data with a matching degree greater than or equal to the preset matching degree are determined from all computing resource demand data; and the computing resource combination that can meet the target computing resource demand data is determined as the target computing resource combination that matches the demand object.

[0110] In this embodiment of the invention, the resource demand prediction data can be used to represent the predicted demand data of the above-mentioned demand objects for computing resources obtained based on the resource collaboration request data, while the computing resource demand data can be used to represent the historical demand data of the corresponding demand objects for computing resources or the actual demand data of the demand objects uploaded to the resource collaboration system in advance.

[0111] In this embodiment of the invention, the demand matching degree is used to represent the degree of matching between the resource demand forecast data and the resource demand data. Optionally, the demand matching degree can be 80%, 75%, or any other preset value; this embodiment of the invention does not impose any limitations.

[0112] As can be seen, this optional embodiment can directly determine from all computing resource combinations the computing resource demand data corresponding to the demand object, and use it as the target computing resource combination that matches the demand object, thereby improving the efficiency and speed of providing the demander with a suitable computing resource combination. Alternatively, based on the resource collaboration request data of the demand object, it can predict the resource demand forecast data corresponding to the demand object, thereby improving the prediction accuracy and efficiency of the resource demand forecast data corresponding to the demand object. Furthermore, based on the resource demand forecast data and computing resource demand data corresponding to the demand object, it can accurately calculate the demand matching degree of the demand object for each computing resource demand data. Based on the accurately calculated demand matching degree, it can quickly and accurately determine the target computing resource demand data with a high demand matching degree from all computing resource demand data, and use its computing resource combination as the target computing resource combination that matches the demand object, thereby improving the matching accuracy and reliability between the demand object and the computing resource combination.

[0113] Example 2

[0114] Please see Figure 2 , Figure 2This is a flowchart illustrating a method for collaborative implementation of AI-based computing resources disclosed in an embodiment of the present invention. Wherein, Figure 2 The described AI-based computing resource collaboration method can be applied to an AI-based computing resource collaboration system. This system may include collaborative devices or collaborative servers, where the collaborative server may be a cloud server or a local server; this embodiment of the invention does not impose limitations. Figure 2 As shown, this AI-based method for collaborative computing resources can include the following operations:

[0115] 201. Receive resource sharing request data submitted by suppliers with computing power resources.

[0116] 202. After verifying the data of the supply object and the resource sharing request submitted by the supply object, add the computing power resources that the supply object can provide to the computing power resource pool.

[0117] 203. Based on the computing power resource demand data corresponding to different demand objects, analyze the performance demand data of each demand object in terms of performance, and based on the computing power resource demand data corresponding to each demand object, analyze the energy efficiency demand data of each demand object in terms of energy efficiency.

[0118] In this embodiment of the invention, optionally, the performance requirement data includes a first sub-requirement data for computing resource utilization and / or a second sub-requirement data for computing-network convergence scheduling delay. The computing-network convergence scheduling delay can be a scheduling delay calculated by computing-network convergence technology, which is used to realize unified management and collaborative scheduling of computing, storage and network resources. This embodiment of the invention does not limit this.

[0119] Optionally, the energy efficiency requirement data may include a third sub-requirement data for the PUE value of the intelligent computing center and / or a fourth sub-requirement data for the liquid cooling technology coverage. The PUE value of the intelligent computing center can represent energy efficiency, which is the ratio of the total power consumption of the data center to the power consumption of IT equipment. The liquid cooling technology coverage can be a coverage rate determined based on efficient heat dissipation technologies such as liquid cooling and air cooling; this embodiment of the invention does not impose any limitations on this.

[0120] 204. Based on the cloud-edge-device collaborative architecture, according to the performance requirement data corresponding to each requirement object, the corresponding nodes are associated in the computing power resource pool to obtain the associated node data of each requirement object in the computing power resource pool.

[0121] In this embodiment of the invention, the associated node data includes edge node data or cloud center data. For example, for objects with low latency and low utilization, edge node data is searched in the computing resource pool and associated; for objects with high latency and high utilization, cloud center data is searched in the computing resource pool and associated.

[0122] 205. Based on the energy efficiency demand data corresponding to each demand object, match each demand object with the computing power resource pool according to different energy efficiency levels of computing power resource areas to obtain the target computing power resource area that matches each demand object.

[0123] In this embodiment of the invention, optionally, the computing resource pool can be divided into a high-efficiency layer computing resource area, a standard layer computing resource area, and a basic layer computing resource area according to energy efficiency. The high-efficiency layer can be used to represent an energy efficiency level with a low PUE value and a high liquid cooling coverage rate, the standard layer can be used to represent an energy efficiency level with a medium PUE value and a low liquid cooling coverage rate, and the basic layer can be used to represent an energy efficiency level with a high PUE value and an almost zero liquid cooling coverage rate. In this way, the computing resource areas of different energy efficiency levels can be quickly and accurately determined from the computing resource pool by using the PUE value of the intelligent computing center and the liquid cooling technology coverage rate, which facilitates the subsequent selection of computing resources that conform to green and low-carbon conditions.

[0124] In this embodiment of the invention, optionally, there is no order between step 205 and step 204. That is, step 205 can occur before step 204, after step 204, or simultaneously with step 204. This embodiment of the invention does not impose any limitations.

[0125] 206. Based on the associated node data of each demand object in the computing power resource pool, and the target computing power resource region that matches the demand object, determine the computing power resource combination that can meet the computing power resource demand data corresponding to the demand object from the computing power resource pool.

[0126] 207. When a resource collaboration request data is received from a certain requesting object, the requesting object is authenticated.

[0127] 208. After the object of the demand is certified, the target computing resource combination that matches the object of the demand is determined from all computing resource combinations based on the computing resource demand data corresponding to the object of the demand.

[0128] 209. Provide computing resources to the target user based on the target computing resource combination.

[0129] In this embodiment of the invention, for other descriptions of steps 201, 202 and steps 207-209, please refer to the detailed description of steps 101, 102 and steps 104-106 in Embodiment 1. These descriptions will not be repeated in this embodiment of the invention.

[0130] It is evident that implementation Figure 2The described AI-based computing resource collaboration method receives resource sharing request data submitted by suppliers with computing resources. After verifying the supplier and its resource sharing request data, the method adds the computing resources that the supplier can provide to a computing resource pool. This allows for the fusion of available computing resources, improving the accuracy and reliability of computing resource sharing. Furthermore, by combining the computing resource demand data of different demanders, the method organizes the computing resources in the pool to form computing resource combinations that meet the diverse computing resource needs. This improves the accuracy of pre-allocation of computing resources and enhances the matching accuracy between demanders and computing resource combinations. Upon receiving and authenticating a resource collaboration request data from a specific demander, the method matches a suitable computing resource combination based on the demander's corresponding computing resource demand data and then provides the computing resources. This pre-allocation of computing resource combinations improves the efficiency, speed, and timeliness of actual computing resource allocation. This not only improves the efficiency and accuracy of providing suitable computing resources to demanders but also enhances resource utilization. Furthermore, in the process of matching suitable computing resource combinations for different needs, by analyzing the performance requirements and energy efficiency requirements of each need, and based on the cloud-edge-device collaborative architecture, corresponding nodes are associated in the computing resource pool according to the performance requirement data of each need. This obtains the associated node data of each need in the computing resource pool, which can improve the accuracy of determining the associated node data of each need based on the performance requirement data. This is conducive to realizing the dynamic allocation and efficient utilization of computing resources based on the cloud-edge-device collaborative framework, thereby improving the utilization rate of computing resources. In addition, based on the energy efficiency requirement data of each need, each need is matched with computing resource areas of different energy efficiency levels divided by the computing resource pool, which obtains the target computing resource area matching each need. This can improve the accuracy of the division of computing resource areas of different energy efficiency levels based on the energy efficiency requirement data, thereby improving the matching accuracy of needs with computing resource areas of different energy efficiency levels, and thus helping to reduce the energy consumption of intelligent computing centers.

[0131] In an optional embodiment, step 206 above, which determines a combination of computing resources that can satisfy the computing resource demand data corresponding to each demand object from the computing resource pool based on the associated node data of each demand object in the computing resource pool and the target computing resource region that matches the demand object, includes:

[0132] Based on the computing power resource demand data corresponding to each demand object, analyze the resource allocation priority data corresponding to each demand object, and generate priority queue data based on all demand objects and the resource allocation priority data corresponding to all demand objects.

[0133] Based on the priority queue data, the target demand objects of the computing resources to be sorted out are identified;

[0134] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource area that matches the target demand object, the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool. Then, the operation of determining the target demand object of the computing power resource to be sorted based on the priority queue data is re-executed until the required computing power resource combination is sorted for all demand objects in the priority queue data.

[0135] In this embodiment of the invention, the resource allocation priority data corresponding to each demand object can be determined by data such as the urgency and importance of the business to be processed by the demand object. For example, demand objects with more urgent and important business have higher priority, while demand objects with relatively low urgency and / or relatively low importance have lower priority.

[0136] For example, based on priority queue data, the highest priority demand object is identified as the target demand object for computing resources to be sorted out. Based on the associated node data of this target demand object in the computing resource pool and the target computing resource region that matches the target demand object, a combination of computing resources that can meet the computing resource demand data corresponding to the target demand object is determined from the computing resource pool. Then, based on the priority queue data again, the second highest priority demand object is identified as the target demand object for computing resources to be sorted out, and the above operation of determining a combination of computing resources that can meet the computing resource demand data corresponding to the target demand object based on the associated node data of this target demand object in the computing resource pool and the target computing resource region that matches the target demand object is repeated. This process is repeated until all the computing resource combinations required by all demand objects in the priority queue data have been sorted out.

[0137] As can be seen, this optional embodiment can accurately analyze the resource allocation priority data corresponding to each demand object based on the computing power resource demand data corresponding to each demand object, and generate priority queue data based on all demand objects and their corresponding resource allocation priority data, thereby improving the accuracy and reliability of priority queue data generation. Subsequently, based on the priority queue data, the target demand object of the computing power resources to be sorted is determined, improving the accuracy of the determination of the target demand object. Then, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource area that matches the target demand object, the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, improving the accuracy and reliability of matching demand objects with computing power resource combinations. Then, the next demand object is determined, and so on, until the queue ends. This is beneficial to improve the accuracy and reliability of determining the computing power resource combination for each demand object by determining the priority of different demand objects, and it is also beneficial to prioritize matching suitable computing power resource combinations for higher priority demand objects when the computing power resource pool is insufficient to meet the needs of all demand objects, thus ensuring that suitable computing power resource combinations are provided for higher priority demand objects.

[0138] In this optional embodiment, as an optional implementation method, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource region that matches the target demand object, a combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, including:

[0139] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource area that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource area. Each computing power resource has corresponding computing power resource data.

[0140] Based on the computing power resource data corresponding to each candidate computing power resource, analyze the scope of computing power service targets for each candidate computing power resource;

[0141] Based on the scope of computing power services of all candidate computing power resources, determine the combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand objects from all candidate computing power resources.

[0142] In this embodiment of the invention, optionally, the scope of computing power service objects can be used to represent the service capability boundary of the corresponding candidate computing power resources, and the scope of computing power service objects can include the types of services that the corresponding candidate computing power resources can serve, and / or the types of users that the corresponding candidate computing power resources can serve. This embodiment of the invention does not impose any limitations.

[0143] For example, for each candidate computing resource, if the scope of the computing service target of the candidate computing resource includes the relevant identification data (such as business type identification data, user type identification data) of the computing resource demand data corresponding to the target demand object, then the candidate computing resource can be assigned to the computing resource combination of the computing resource demand data corresponding to the target demand object.

[0144] As can be seen, the optional implementation can determine all candidate computing resources matching the target demand object from all computing resources in the target computing resource region based on the associated node data of the target demand object in the computing resource pool and the target computing resource region that matches the target demand object. It can improve the accuracy and efficiency of determining the candidate computing resources matching the target demand object based on the associated node data and the target computing resource region. Furthermore, based on the computing resource data corresponding to each candidate computing resource, it can accurately analyze the computing power service object range of each candidate computing resource. Thus, based on the computing power service object range of all candidate computing resources, it can determine the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object from all candidate computing resources. This can improve the accuracy and reliability of determining the computing power resource combination corresponding to the computing power demand data of the target demand object.

[0145] In this optional implementation, optionally, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource region that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource region, including:

[0146] Based on the computing power resource requirements data corresponding to the target demand object, determine the expected computing power deployment range corresponding to the target demand object;

[0147] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all computing power resources under the expected computing power deployment range are obtained from all computing power resources in the target computing power resource region, as all alternative computing power resources that match the target demand object.

[0148] As can be seen, this optional implementation can also accurately determine the expected computing power deployment range corresponding to the target demand object based on the computing power resource demand data corresponding to the target demand object. Thus, based on the associated node data and the target computing power resource region, it can obtain all computing power resources under the expected computing power deployment range from all computing power resources in the target computing power resource region, as all candidate computing power resources matching the target demand object. This improves the accuracy and reliability of obtaining candidate computing power resources for determining the computing power resource combination, thereby helping to further improve the accuracy and reliability of subsequent matching of demand objects and computing power resource combinations.

[0149] In another optional embodiment, the cloud-edge-device collaborative architecture in step 204 above, based on the performance requirement data corresponding to each demand object, associates the corresponding nodes in the computing resource pool to obtain the associated node data of each demand object in the computing resource pool, including:

[0150] Based on the computing power resource demand data corresponding to each demand object, determine the first demand level for the first sub-demand data and the second demand level for the second sub-demand data respectively.

[0151] Based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, the corresponding node is associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, the corresponding node is associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data.

[0152] For each demand object, the associated node data of the demand object in the computing power resource pool is determined based on the first node data corresponding to the demand object, the first degree of demand for the first node data, the second node data corresponding to the demand object, and the second degree of demand for the second node data.

[0153] In this embodiment of the invention, specifically, for each demand object, the first degree of demand for the first node data and the second degree of demand for the second node data are compared to obtain the demand degree comparison result;

[0154] When the demand level comparison result indicates that the absolute value of the difference between the first demand level and the second demand level is greater than or equal to the preset difference, the node data corresponding to the maximum demand level between the first demand level and the second demand level is determined as the associated node data of the demand object in the computing power resource pool.

[0155] When the comparison result of the demand level indicates that the absolute value of the difference between the first demand level and the second demand level is less than the preset difference, the overlapping node data in the first node data and the second node data is determined as the associated node data of the demand object in the computing power resource pool.

[0156] As can be seen, this optional embodiment can determine the first degree of demand for the first sub-demand data and the second degree of demand for the second sub-demand data based on the computing power resource demand data corresponding to each demand object, thereby improving the accuracy and efficiency of determining the degree of demand corresponding to different demand data at the performance level. Subsequently, based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, corresponding nodes are associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, corresponding nodes are associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data, thereby improving the accuracy of obtaining the node data associated with different demand data at the performance level. Thus, for each demand object, based on the first node data corresponding to the demand object, the first degree of demand for the first node data, the second node data corresponding to the demand object, and the second degree of demand for the second node data, the associated node data of the demand object in the computing power resource pool is determined, thereby improving the accuracy and reliability of determining the associated node data of the demand object in the computing power resource pool.

[0157] Example 3

[0158] Please see Figure 3 , Figure 3 This is a schematic diagram of a computing resource collaborative implementation system based on AI, as disclosed in an embodiment of the present invention. Figure 3 The described AI-based computing resource collaborative implementation system may include collaborative devices or collaborative servers, wherein the collaborative server may include a cloud server or a local server, and the embodiments of the present invention are not limited thereto. Figure 3 As shown, this AI-based computing resource collaborative implementation system may include:

[0159] The data receiving module 301 is used to receive resource sharing request data submitted by the supply object with computing power resources. The resource sharing request data is used to apply for computing power resources to be provided to the pre-built computing power resource pool for scheduling by the computing power resource pool.

[0160] The resource sharing module 302 is used to add the computing power resources that the supply object can provide to the computing power resource pool after verifying the supply object and the resource sharing request data submitted by the supply object.

[0161] The resource sorting module 303 is used to sort out the computing resources in the computing resource pool based on the computing resource demand data corresponding to different demand objects, so as to determine the computing resource combination that can meet the computing resource demand data of different computing resource objects.

[0162] The data authentication module 304 is used to authenticate the resource collaboration request data of a certain request object when it receives the resource collaboration request data of a certain request object.

[0163] The resource matching module 305 is used to determine the target computing resource combination that matches the demand object from all computing resource combinations after the demand object has been authenticated.

[0164] The resource scheduling module 306 is used to provide computing resources to the target object based on the target computing resource combination.

[0165] It is evident that implementation Figure 3 The described AI-based computing resource collaboration system can receive resource sharing request data submitted by suppliers with computing resources. After verifying the supplier and its resource sharing request data, the system adds the computing resources that the supplier can provide to the computing resource pool. This allows for the fusion of available computing resources, improving the accuracy and reliability of computing resource sharing. Furthermore, by combining the computing resource demand data of different demanders, the system organizes the computing resources in the pool to form computing resource combinations that meet the diverse computing resource needs. This improves the accuracy of pre-allocation of computing resources and enhances the matching accuracy between demanders and computing resource combinations. Upon receiving and authenticating a resource collaboration request data from a specific demander, the system matches a suitable computing resource combination based on the demander's corresponding computing resource demand data and then provides the computing resources. This pre-allocation of computing resource combinations improves the efficiency, speed, and timeliness of actual computing resource allocation. This not only improves the efficiency and accuracy of providing suitable computing resources to demanders but also enhances resource utilization.

[0166] In an optional embodiment, the resource allocation module 303, based on the acquired computing resource demand data corresponding to different demand objects, allocates computing resources in the computing resource pool to determine the specific methods for combining computing resources that can meet the different computing resource demand data, including:

[0167] Based on the computing power resource demand data corresponding to different demand objects, analyze the performance demand data of each demand object in terms of performance, and based on the computing power resource demand data corresponding to each demand object, analyze the energy efficiency demand data of each demand object in terms of energy efficiency.

[0168] Based on the cloud-edge-device collaborative architecture, according to the performance requirement data corresponding to each demand object, the corresponding nodes are associated in the computing power resource pool to obtain the associated node data of each demand object in the computing power resource pool. The associated node data includes edge node data or cloud center data.

[0169] Based on the energy efficiency requirement data corresponding to each demand object, each demand object is matched with the computing power resource area of ​​different energy efficiency levels obtained by dividing the computing power resource pool according to energy efficiency, so as to obtain the target computing power resource area that matches each demand object.

[0170] Based on the associated node data of each demand object in the computing power resource pool, and the target computing power resource region that matches the demand object, a combination of computing power resources that can meet the computing power resource demand data corresponding to the demand object is determined from the computing power resource pool.

[0171] As can be seen, this optional embodiment can, in the process of matching suitable computing resource combinations for different needs, analyze the performance requirements data of each need object at the performance level and the energy efficiency requirements data at the energy efficiency level. Based on the cloud-edge-device collaborative architecture, it associates corresponding nodes in the computing resource pool according to the performance requirement data of each need object, thereby obtaining the associated node data of each need object in the computing resource pool. This can improve the accuracy of determining the associated node data of each need object based on the performance requirement data, which is conducive to realizing the dynamic allocation and efficient utilization of computing resources based on the cloud-edge-device collaborative framework, thereby improving the utilization rate of computing resources. Furthermore, based on the energy efficiency requirement data of each need object, it matches each need object with computing resource areas of different energy efficiency levels divided by the computing resource pool according to energy efficiency, thereby obtaining the target computing resource area that matches each need object. This can improve the accuracy of dividing computing resource areas of different energy efficiency levels based on the energy efficiency requirement data, thereby improving the matching accuracy of need objects with computing resource areas of different energy efficiency levels, and thus helping to reduce the energy consumption of intelligent computing centers.

[0172] In this optional embodiment, as an optional implementation method, the resource sorting module 303 determines, based on the associated node data of each demand object in the computing power resource pool and the target computing power resource region matching the demand object, the specific method for determining the computing power resource combination that can meet the computing power resource demand data corresponding to the demand object from the computing power resource pool includes:

[0173] Based on the computing power resource demand data corresponding to each demand object, analyze the resource allocation priority data corresponding to each demand object, and generate priority queue data based on all demand objects and the resource allocation priority data corresponding to all demand objects.

[0174] Based on the priority queue data, the target demand objects of the computing resources to be sorted out are identified;

[0175] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource area that matches the target demand object, the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool. Then, the operation of determining the target demand object of the computing power resource to be sorted based on the priority queue data is re-executed until the required computing power resource combination is sorted for all demand objects in the priority queue data.

[0176] As can be seen, this optional implementation can accurately analyze the resource allocation priority data corresponding to each demand object based on the computing power resource demand data corresponding to each demand object, and generate priority queue data based on all demand objects and their corresponding resource allocation priority data, thereby improving the accuracy and reliability of priority queue data generation. Subsequently, based on the priority queue data, the target demand object of the computing power resources to be sorted is determined, improving the accuracy of the determination of the target demand object. Then, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource area that matches the target demand object, the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, improving the accuracy and reliability of matching demand objects with computing power resource combinations. Then, the next demand object is determined, and so on, until the queue ends. This is beneficial to improve the accuracy and reliability of determining the computing power resource combination for each demand object by determining the priority of different demand objects, and it is also beneficial to prioritize matching suitable computing power resource combinations for higher priority demand objects when the computing power resource pool is insufficient to meet the needs of all demand objects, thus ensuring that suitable computing power resource combinations are provided for higher priority demand objects.

[0177] In this optional implementation, the resource allocation module 303 may determine, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource region that matches the target demand object, the specific method for determining the computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object from the computing power resource pool includes:

[0178] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource area that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource area. Each computing power resource has corresponding computing power resource data.

[0179] Based on the computing power resource data corresponding to each candidate computing power resource, analyze the scope of computing power service targets for each candidate computing power resource;

[0180] Based on the scope of computing power services of all candidate computing power resources, determine the combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand objects from all candidate computing power resources.

[0181] As can be seen, this optional implementation can also determine all candidate computing resources matching the target demand object from all computing resources in the target computing resource region based on the associated node data of the target demand object in the computing resource pool and the target computing resource region that matches the target demand object. It can improve the accuracy and efficiency of determining the candidate computing resources matching the target demand object based on the associated node data and the target computing resource region. Furthermore, it can accurately analyze the computing service object range of each candidate computing resource based on the computing resource data corresponding to each candidate computing resource. Thus, based on the computing service object range of all candidate computing resources, it can determine the combination of computing resources that can meet the computing resource demand data corresponding to the target demand object from all candidate computing resources. This can improve the accuracy and reliability of determining the combination of computing resources corresponding to the computing resource demand data of the target demand object.

[0182] In this optional implementation, further optionally, the resource sorting module 303 determines all candidate computing resources matching the target demand object from all computing resources in the target computing resource region based on the associated node data of the target demand object in the computing resource pool and the target computing resource region that matches the target demand object. Specifically, this includes:

[0183] Based on the computing power resource requirements data corresponding to the target demand object, determine the expected computing power deployment range corresponding to the target demand object;

[0184] Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all computing power resources under the expected computing power deployment range are obtained from all computing power resources in the target computing power resource region, as all alternative computing power resources that match the target demand object.

[0185] As can be seen, this optional implementation can also accurately determine the expected computing power deployment range corresponding to the target demand object based on the computing power resource demand data corresponding to the target demand object. Thus, based on the associated node data and the target computing power resource region, it can obtain all computing power resources under the expected computing power deployment range from all computing power resources in the target computing power resource region, as all candidate computing power resources matching the target demand object. This improves the accuracy and reliability of obtaining candidate computing power resources for determining the computing power resource combination, thereby helping to further improve the accuracy and reliability of subsequent matching of demand objects and computing power resource combinations.

[0186] In this optional embodiment, as another optional implementation, the performance requirement data includes first sub-requirement data for computing resource utilization and / or second sub-requirement data for computing-network converged scheduling latency. Furthermore, the resource management module 303, based on a cloud-edge-device collaborative architecture, associates corresponding nodes in the computing resource pool according to the performance requirement data corresponding to each requirement object. The specific method for obtaining the associated node data of each requirement object in the computing resource pool includes:

[0187] Based on the computing power resource demand data corresponding to each demand object, determine the first demand level for the first sub-demand data and the second demand level for the second sub-demand data respectively.

[0188] Based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, the corresponding node is associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, the corresponding node is associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data.

[0189] For each demand object, the associated node data of the demand object in the computing power resource pool is determined based on the first node data corresponding to the demand object, the first degree of demand for the first node data, the second node data corresponding to the demand object, and the second degree of demand for the second node data.

[0190] As can be seen, this optional implementation can determine the first degree of demand for the first sub-demand data and the second degree of demand for the second sub-demand data based on the computing power resource demand data corresponding to each demand object, thereby improving the accuracy and efficiency of determining the degree of demand corresponding to different demand data at the performance level. Subsequently, based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, corresponding nodes are associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, corresponding nodes are associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data, thereby improving the accuracy of obtaining the node data associated with different demand data at the performance level. Thus, for each demand object, based on the first node data corresponding to the demand object, the first degree of demand for the first node data, the second node data corresponding to the demand object, and the second degree of demand for the second node data, the associated node data of the demand object in the computing power resource pool is determined, thereby improving the accuracy and reliability of determining the associated node data of the demand object in the computing power resource pool.

[0191] In another optional embodiment, the resource matching module 305 determines the target computing resource combination that matches the demand object from all computing resource combinations based on the computing resource demand data corresponding to the demand object. Specifically, this includes:

[0192] From all available computing resource combinations, identify the computing resource combination that can meet the computing resource requirements of the target user, and use this combination as the target computing resource combination that matches the target user's requirements; or...

[0193] Based on the resource collaboration request data of the demand object, predict the resource demand forecast data corresponding to the demand object;

[0194] Based on the resource demand forecast data corresponding to the demand object and the computing power resource demand data for each demand object, calculate the demand matching degree of the demand object with respect to the computing power resource demand data for each demand object.

[0195] Based on the matching degree of this demand with all computing resource demand data, target computing resource demand data with a matching degree greater than or equal to a preset matching degree is identified from all computing resource demand data; and the computing resource combinations that can meet the target computing resource demand data are identified as target computing resource combinations that match this demand object.

[0196] As can be seen, this optional embodiment can directly determine from all computing resource combinations the computing resource demand data corresponding to the demand object, and use it as the target computing resource combination that matches the demand object, thereby improving the efficiency and speed of providing the demander with a suitable computing resource combination. Alternatively, based on the resource collaboration request data of the demand object, it can predict the resource demand forecast data corresponding to the demand object, thereby improving the prediction accuracy and efficiency of the resource demand forecast data corresponding to the demand object. Furthermore, based on the resource demand forecast data and computing resource demand data corresponding to the demand object, it can accurately calculate the demand matching degree of the demand object for each computing resource demand data. Based on the accurately calculated demand matching degree, it can quickly and accurately determine the target computing resource demand data with a high demand matching degree from all computing resource demand data, and use its computing resource combination as the target computing resource combination that matches the demand object, thereby improving the matching accuracy and reliability between the demand object and the computing resource combination.

[0197] Example 4

[0198] Please see Figure 4 , Figure 4 This is a schematic diagram of another AI-based computing resource collaborative implementation system disclosed in an embodiment of the present invention. Figure 4 As shown, this AI-based computing resource collaborative implementation system may include:

[0199] Memory 401 storing executable program code;

[0200] Processor 402 coupled to memory 401;

[0201] The processor 402 calls the executable program code stored in the memory 401 to execute some or all of the steps in the AI-based computing resource collaborative implementation method described in Embodiment 1 or Embodiment 2 of the present invention.

[0202] Example 5

[0203] This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the AI-based computing resource collaborative implementation methods disclosed in Embodiment 1 of this invention.

[0204] Example 6

[0205] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform some or all of the steps in the AI-based computing resource collaborative implementation method described in Embodiment 1 or Embodiment 2.

[0206] The system embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0207] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0208] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collaborative implementation of computing resources based on AI, characterized in that, The method includes: Receive resource sharing request data submitted by a supply object with computing power resources. The resource sharing request data is used to request the provision of computing power resources to a pre-built computing power resource pool for scheduling by the computing power resource pool. After the supply object and the resource sharing request data submitted by the supply object are verified, the computing power resources that the supply object can provide are added to the computing power resource pool. Based on the computing power resource demand data corresponding to different demand objects, the computing power resources in the computing power resource pool are sorted out to determine the computing power resource combination that can meet the different computing power resource demand data. When a resource collaboration request data of a certain requested object is received, the requested object is authenticated; After the object of the demand is authenticated, a target computing resource combination that matches the object of the demand is determined from all the computing resource combinations based on the computing resource demand data corresponding to the object of the demand; and computing resources are provided to the object of the demand based on the target computing resource combination. The step of sorting out the computing resources in the computing resource pool based on the computing resource demand data corresponding to different demand objects to determine the computing resource combination that can meet the different computing resource demand data includes: Based on the obtained computing resource demand data corresponding to different demand objects, analyze the performance demand data of each demand object at the performance level, and based on the computing resource demand data corresponding to each demand object, analyze the energy efficiency demand data of each demand object at the energy efficiency level. The performance demand data includes the first sub-demand data for computing resource utilization and the second sub-demand data for computing network convergence scheduling latency. The energy efficiency demand data includes the third sub-demand data for the PUE value of the intelligent computing center and the fourth sub-demand data for the coverage of liquid cooling technology. Based on the cloud-edge-device collaborative architecture, according to the performance requirement data corresponding to each of the aforementioned demand objects, corresponding nodes are associated in the computing power resource pool to obtain the associated node data of each of the aforementioned demand objects in the computing power resource pool. The associated node data includes edge node data or cloud center data. Based on the energy efficiency requirement data corresponding to each of the aforementioned requirement objects, each of the aforementioned requirement objects is matched with the computing power resource areas of different energy efficiency levels obtained by dividing the computing power resource pool according to energy efficiency, so as to obtain the target computing power resource area that matches each of the aforementioned requirement objects. Based on the associated node data of each demand object in the computing power resource pool, and the target computing power resource region that matches the demand object, a computing power resource combination that can meet the computing power resource demand data corresponding to the demand object is determined from the computing power resource pool. The step of determining, based on the associated node data of each demand object in the computing resource pool and the target computing resource region matching the demand object, a combination of computing resources that can satisfy the computing resource demand data corresponding to the demand object is determined from the computing resource pool, including: Based on the computing power resource demand data corresponding to each demand object, analyze the resource allocation priority data corresponding to each demand object, and generate priority queue data based on all demand objects and the resource allocation priority data corresponding to all demand objects. Based on the priority queue data, the target demand objects of the computing resources to be sorted out are determined; Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource area that matches the target demand object, a computing power resource combination that can meet the computing power resource demand data corresponding to the target demand object is determined from the computing power resource pool, and the operation of determining the target demand object of the current computing power resource to be sorted based on the priority queue data is re-executed until the required computing power resource combination is sorted for all demand objects in the priority queue data. And, the step of determining, based on the associated node data of the target demand object in the computing power resource pool and the target computing power resource region matching the target demand object, a combination of computing power resources that can satisfy the computing power resource demand data corresponding to the target demand object from the computing power resource pool includes: Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all candidate computing power resources that match the target demand object are determined from all computing power resources in the target computing power resource region, and each computing power resource has corresponding computing power resource data; Based on the computing power resource data corresponding to each of the candidate computing power resources, analyze the scope of computing power service targets for each of the candidate computing power resources; Based on the scope of computing power service targets of all the candidate computing power resources, determine the combination of computing power resources that can meet the computing power resource demand data corresponding to the target demand object from all the candidate computing power resources.

2. The AI-based computing resource collaborative implementation method according to claim 1, characterized in that, The step of determining all candidate computing resources matching the target demand object from all computing resources in the target computing resource region, based on the associated node data of the target demand object in the computing resource pool and the target computing resource region that matches the target demand object, includes: Based on the computing power resource demand data corresponding to the target demand object, determine the expected computing power deployment range corresponding to the target demand object; Based on the associated node data of the target demand object in the computing power resource pool, and the target computing power resource region that matches the target demand object, all computing power resources under the desired computing power deployment range are obtained from all computing power resources in the target computing power resource region, as all alternative computing power resources that match the target demand object.

3. The AI-based computing resource collaborative implementation method according to claim 1 or 2, characterized in that, The cloud-edge-device collaborative architecture, based on the performance requirement data corresponding to each of the aforementioned demand objects, associates corresponding nodes in the computing resource pool to obtain the associated node data of each demand object in the computing resource pool, including: Based on the computing power resource demand data corresponding to each demand object, determine the first demand level for the first sub-demand data and the second demand level for the second sub-demand data respectively. Based on the cloud-edge-device collaborative architecture, according to the first sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the first node data associated with the first sub-demand data; and according to the second sub-demand data corresponding to each demand object, a corresponding node is associated in the computing power resource pool to obtain the second node data associated with the second sub-demand data. For each of the aforementioned demand objects, the associated node data of the demand object in the computing power resource pool is determined based on the first node data corresponding to the demand object, the first demand level for the first node data, the second node data corresponding to the demand object, and the second demand level for the second node data.

4. The AI-based computing resource collaborative implementation method according to claim 1 or 2, characterized in that, The step of determining the target computing resource combination that matches the demand object from all the computing resource combinations based on the computing resource demand data corresponding to the demand object includes: From all the aforementioned computing resource combinations, determine the computing resource combination that can meet the computing resource requirement data corresponding to the object with the requirement, and use it as the target computing resource combination matching the object with the requirement; or... Based on the resource collaboration request data of the demand object, predict the resource demand forecast data corresponding to the demand object; Based on the resource demand forecast data corresponding to the demand object and the computing power resource demand data for each of the aforementioned demand objects, calculate the demand matching degree of the demand object with respect to each of the aforementioned computing power resource demand data. Based on the matching degree of the demand object with all the computing power resource demand data, target computing power resource demand data with a matching degree greater than or equal to a preset matching degree are determined from all the computing power resource demand data; and the computing power resource combination that can meet the target computing power resource demand data is determined as the target computing power resource combination that matches the demand object.

5. A system for collaborative realization of computing resources based on AI, characterized in that, The system is used to execute the AI-based computing resource collaborative implementation method as described in any one of claims 1-4, and the system comprises: The data receiving module is used to receive resource sharing request data submitted by a supply object with computing power resources. The resource sharing request data is used to request computing power resources to be provided to a pre-built computing power resource pool for scheduling by the computing power resource pool. The resource sharing module is used to add the computing power resources that the supply object can provide to the computing power resource pool after the verification of the supply object and the resource sharing request data submitted by the supply object is passed. The resource sorting module is used to sort out the computing resources in the computing resource pool based on the computing resource demand data corresponding to different demand objects, so as to determine the computing resource combination that can meet the computing resource demand data of different computing resource objects. The data authentication module is used to authenticate the requesting object when it receives resource collaboration request data from a certain requesting object. The resource matching module is used to determine the target computing resource combination that matches the demand object from all the computing resource combinations after the demand object has been authenticated, based on the computing resource demand data corresponding to the demand object. The resource scheduling module is used to provide computing resources to the object with the demand based on the target computing resource combination.

6. A system for collaborative realization of computing resources based on AI, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the AI-based computing resource collaborative implementation method as described in any one of claims 1-4.

7. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the AI-based computing resource collaborative implementation method as described in any one of claims 1-4.

Images