A computing power service platform and method

Abstract

The application provides a computing power service platform and method, the platform comprises a computing power request issuing component, a computing node determining component and an association establishing component. The method comprises obtaining government, enterprise and scientific research institution and other user types of computing power service, giving different computing power service paths to different user types; issuing a computing power request instruction according to the computing power service path; decomposing the computing power service request instruction according to the computing power request instruction to obtain a plurality of sub-problems corresponding to the computing power service request, determining at least one target computing power computing node for each sub-problem; sending a computing power control instruction to the target computing power computing node, receiving the computing result returned by the target computing power computing node, and establishing the association between the target computing power computing node and the computing power service demand. The application has a hierarchical scheduling capability that adapts to needs, and automatically matches different computing power service paths for different user types such as government, enterprise and scientific research institution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information and data processing, and particularly relates to a computing power service platform and method. BACKGROUND

[0002] With the rapid development of technologies such as artificial intelligence, the Internet of Things and the meta-universe, the global computing power demand is growing exponentially, but the existing computing power service system faces the following core contradictions: resource heterogeneity and fragmentation: computing power resources are scattered in different architectures (CPU / GPU / FPGA), different service providers (public cloud, private cloud, edge node) and different geographical locations, lacking unified access standards and coordination mechanisms, resulting in less than 30% resource utilization. Service mode is rigid: traditional computing power supply is mainly based on static allocation (such as virtual machine rental), which cannot dynamically adapt to real-time changes in computing tasks (such as sudden AI inference and streaming data processing), and the task queuing delay is as high as minutes. Lack of transaction trust: the computing power transaction process lacks transparency and verifiability, and it is difficult for both parties to reach a credible consensus on service quality (such as latency and accuracy), with a dispute rate of more than 15%. Security and compliance risks: computing power resources are not completely isolated in a multi-tenant environment, sensitive data may be leaked through shared hardware side channels, and cross-border computing power scheduling faces data sovereignty compliance challenges. The above problems lead to an imbalance between "local overload and global idling" of computing power resources, and an integrated service platform that can realize cross-domain interconnection, intelligent scheduling and trusted transactions is urgently needed.

[0003] Prior art one, application number: CN202211462571.8 discloses a computing power service data set storage method, a computing power computing device and a computing power service equipment, which comprises: a control module determines that the computing power service data set needs to be updated, if the memory capacity of the computing module is greater than or equal to the storage capacity required by the updated second data set but less than the total storage capacity required by the updated first data set and the second data set, then the new first data set and the second data set are generated, the new first data set is saved in the storage module, and the new second data set is sent to the computing module and saved in the memory of the computing module; wherein the second data set is generated from the first data set and directly participates in the execution of the computing module to perform computing tasks, and the first data set does not directly participate in the execution of the computing module to perform computing tasks. Although it can prolong the memory usage time of the computing module and the service life of the entire computing power computing device; but the computing data set (first / second data set) needs to be split, which limits the dynamic expansion of memory capacity and cannot adapt to sudden computing power demand.

[0004] The prior art two, application number: CN202410581018.9 discloses a kind of computing power service security capability grade evaluation method, system and computer equipment, method includes: obtaining the computing power service system of the enterprise to be evaluated and the evaluation demand corresponding to computing power service system;According to the evaluation demand corresponding to computing power service system, the first-level security evaluation index set in the computing power security capability evaluation architecture is clipped, to obtain effective first-level security evaluation index;According to the evaluation demand corresponding to computing power service system, the second-level security evaluation index in effective first-level security evaluation index is clipped, to obtain effective second-level security evaluation index;According to effective second-level security evaluation index, computing power service system is evaluated, and the first-level index score of effective first-level security evaluation index corresponding to effective second-level security evaluation index is obtained;According to the first-level index score of effective first-level security evaluation index, the security capability grade of computing power service system is determined.Although it can improve the universality of evaluation method;But through the clipping evaluation index realizes universality, but cannot dynamically perceive computing power environment risk.

[0005] The prior art three, application number: CN 202110335831.4 discloses a kind of computing power service authentication method, equipment and readable storage medium, to solve the problem of lower security of existing computing power service.The method includes: first, computing power service authentication device receives the first access request sent by computing power providing device, and the first access request is used to request access to computing power network, and the first access request includes idle resource information of computing power providing device.Then, computing power service authentication device determines target test task according to idle resource information, and target test task is the test task corresponding to idle resource information and sends target test task to computing power providing device.Finally, computing power service authentication device receives target data sent by computing power providing device within preset time, and sends first indication information to computing power providing device, target data is the correct result of target test task, and first indication information is used to indicate that the device has passed the access authentication of computing power network.Although computing power provider receives resource rental fee paid by computing power demander, and informs other devices in computing power network through broadcast, thereby increasing the publicity of computing power network, and further improving the security of computing power network;But only through test task to verify the basic ability of equipment, but cannot guarantee the stability of continuous service.

[0006] At present, the prior art one, the prior art two and the prior art three have the problem of accurate matching and trusted cooperation of multi-modal computing power resources in dynamic uncertain environment.Therefore, the present application provides a kind of computing power service platform and method. SUMMARY

[0007] To solve the above technical problems, the present application provides a kind of computing power service platform, including:

[0008] The computing power request issuing component is responsible for obtaining the user types of government, enterprises and scientific research institutions of the computing power service, and giving different computing power service paths to different user types; when a user issues a computing power service demand, the corresponding computing power service path is obtained according to the user type, and a computing power request instruction is issued according to the computing power service path;

[0009] The computing node determination component is responsible for decomposing the computing power service request instruction according to the computing power request instruction, obtaining a plurality of sub-problems corresponding to the computing power service request, and corresponding target algorithm types and computing modes;

[0010] The correlation relationship establishment component is responsible for sending a computing power control instruction to the target computing power computing node, receiving the computing result returned by the target computing power computing node, and establishing the correlation relationship between the target computing power computing node and the computing power service demand.

[0011] Optionally, the computing power request issuing component comprises:

[0012] The type confirmation module is responsible for obtaining the computing power service request and the corresponding input mode, triggering a user type identification program by the computing power service request, determining the user type of the computing power service request according to the input mode, and determining the performance interval of the required computing power service path according to the user type;

[0013] The path search module is responsible for constructing a computing power service path database corresponding to different user types, and determining the computing power service path corresponding to the user of the input computing power service request by taking the performance interval as an index of the computing power service path database;

[0014] The instruction conversion module is responsible for obtaining a computing power request instruction in response to the computing power service request, and the computing power request instruction comprises the user type and the corresponding computing power service path.

[0015] Optionally, the path search module comprises:

[0016] The performance value definition submodule is responsible for establishing the computing power service path database, and each demand computing power node and response computing power node is provided with a performance value;

[0017] The path end value confirmation submodule is responsible for querying the target demand computing power node corresponding to the minimum performance value of the performance interval as the starting point of the computing power service path when the index of the performance interval is received, and querying the target response computing power node corresponding to the maximum performance value of the performance interval as the terminal point of the computing power service path;

[0018] The path output submodule is responsible for connecting the target demand computing power node and the target response computing power node to obtain the computing power service path, i.e., determining the computing power service path corresponding to the user of the input computing power service request.

[0019] Optionally, the computing power service path database comprises a demand computing power computing node table, a response computing power computing node table, a computing power service path table, and a performance interval index computing power service path database.

[0020] Optionally, the computing node determination component comprises:

[0021] The task decomposition module is responsible for receiving the computing power request instruction, calling the instruction decomposition program, task decomposition of the computing power request instruction, obtaining the task required by the computing power request instruction, and the task comprising a sub-problem, a target algorithm type, and an algorithm mode.

[0022] The task establishment module is responsible for establishing a computing task corresponding to the sub-problem according to the matching relationship of the target algorithm type and the algorithm mode, and screening at least one computing task from a plurality of computing tasks.

[0023] The node matching module is responsible for calculating the computing power of each computing power computing node comprising processing speed and parallel processing capability, matching the computing demand in the sub-problem corresponding computing task with the computing power, obtaining the computing power computing node meeting the computing power, and determining the target computing power computing node processing the sub-problem by using the task scheduling strategy.

[0024] Optionally, the task decomposition module comprises:

[0025] The program content setting submodule is responsible for taking the computing power request instruction as a trigger event of the instruction decomposition program, for task decomposition of the computing power request instruction, and the running instruction decomposition program is obtained by compiling and executing the first code of the instruction decomposition program.

[0026] The decomposition list generation submodule is responsible for executing the first code, displaying the task in the computing power request instruction, decomposing the computing power service request into a time node of a sub-problem, assigning the task order according to the time node sequence, and generating a task decomposition list.

[0027] The task classification integration submodule is responsible for loading the task attribute in the task decomposition list, calling the second code to classify and integrate the task attribute, generating a task decomposition result set, and obtaining the sub-problem, the target algorithm type, and the algorithm mode under different time nodes.

[0028] Optionally, the first code comprises a calling code of a candidate running code in the second code, the program code of the instruction decomposition program comprises the first code and the second code, the first part of the code is a code for compiling and displaying the task in the computing power request instruction, and the second code is a code for classifying and integrating the task attribute.

[0029] Optionally, the node matching module comprises:

[0030] The index data calculation submodule is responsible for acquiring the computing power resource index data of all computing power calculation nodes, including processor use efficiency, memory usage and load rate, etc.

[0031] The prediction data calculation submodule is responsible for constructing a computing power calculation correlation graph according to the computing power resource index data, and predicting computing power demand prediction data according to the computing demand in the corresponding computing task of the subproblem.

[0032] The resource score calculation submodule is responsible for obtaining the computing power resource score of the computing power calculation node based on the computing power calculation correlation graph and the computing power demand prediction data, and obtaining the computing power calculation node that meets the computing power.

[0033] Optionally, the computing node determination component determines at least one computing task corresponding to each subproblem, and determines at least one target computing power calculation node for each subproblem according to the computing power calling relationship of the computing power calculation node.

[0034] The application provides a computing power service method, which comprises the following steps:

[0035] The types of government, enterprise and scientific research institution users of the computing power service are acquired, different computing power service paths are given to different user types, when the user issues a computing power service demand, the corresponding computing power service path is obtained according to the user type, and a computing power request instruction is issued according to the computing power service path;

[0036] According to the computing power request instruction, the computing power service request instruction is decomposed to obtain a plurality of subproblems, corresponding target algorithm types and computing modes corresponding to the computing power service request, at least one computing task corresponding to each subproblem is determined, and at least one target computing power calculation node for each subproblem is determined according to the computing power calling relationship of the computing power calculation node.

[0037] The computing power control instruction is sent to the target computing power calculation node, the computing result returned by the target computing power calculation node is received, and the association relationship between the target computing power calculation node and the computing power service demand is established.

[0038] The on-demand adaptive hierarchical scheduling capability of the application automatically matches the differentiated computing power service paths for different user types such as governments, enterprises and scientific research institutions (for example, government needs may be biased towards stability and security, and scientific research needs pay more attention to high-performance computing); can quickly locate the optimal service mode according to the user identity, and avoid resource waste or insufficient computing power. Task intelligent decomposition and efficient scheduling, when large computing demand enters the system, the platform can automatically analyze into multiple parallel computing sub-tasks; based on the characteristics of the sub-tasks (algorithm type, calculation mode), flexibly match the appropriate computing nodes (such as GPU cluster suitable for deep learning, CPU cluster suitable for traditional numerical calculation), form a temporary computing alliance. The traceable full-link management ensures that each computing link is controllable and traceable through the issuance of computing power instructions and the recovery of results; bind the calculation results with the original demand, which guarantees the task integrity.

[0039] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims.

[0040] The technical solutions of the application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0041] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, together with the embodiments of the application, to explain the application, and do not constitute a limitation on the application. In the drawings:

[0042] Figure 1 The figure is the block diagram of the computing power service platform in embodiment 1 of the application;

[0043] Figure 2 The figure is the block diagram of the computing power request issuing component in embodiment 2 of the application;

[0044] Figure 3 The figure is the block diagram of the computing node determination component in embodiment 4 of the application;

[0045] Figure 4 The figure is the flow chart of the computing power service method in embodiment 7 of the application. DETAILED DESCRIPTION

[0046] The preferred embodiments of the application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and do not limit the application.

[0047] The terminology used in the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the embodiments of the present application. As used in the description of the embodiments of the present application, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It also will be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0048] The following description refers to the accompanying drawings. Wherever possible, the same reference numbers in different drawings refer to the same or similar elements. The following exemplary embodiments described are not meant to be limiting of all embodiments that can be consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application. In the description of the embodiments of the present application, it will be understood that the terms "first," "second," "third," etc. are used merely as labels, and are not intended to signify sequentially or numerically, unless specifically indicated. The terms "first," "second," "third," etc. are not intended to denote or be related to importance unless the context clearly indicates otherwise.

[0049] Embodiment 1: As shown in the figure, the embodiments of the present application provide a computing power service platform, comprising: Figure 1

[0050] A computing power request issuing component is responsible for obtaining user types of governments, enterprises and scientific research institutions, etc. of computing power services, and giving different computing power service paths to different user types; when a user issues a computing power service demand, a corresponding computing power service path is obtained according to the user type, and a computing power request instruction is issued according to the computing power service path;

[0051] A computing node determination component is responsible for decomposing the computing power service request instruction according to the computing power request instruction, obtaining a plurality of sub-problems corresponding to the computing power service request, and corresponding target algorithm types and computing modes, determining at least one computing task corresponding to each sub-problem, and determining at least one target computing power computing node for each sub-problem according to the computing power computing node computing power calling relationship;

[0052] An association relationship establishment component is responsible for sending a computing power control instruction to the target computing power computing node, receiving a computing result returned by the target computing power computing node, and establishing an association relationship between the target computing power computing node and the computing power service demand.

[0053] ​The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the computing power request issuing component obtains user types such as government agencies, enterprises, and research institutions that provide computing power services, and assigns different computing power service paths to different user types. When a user issues a computing power service request, the component obtains the corresponding computing power service path based on the user type and issues a computing power request instruction based on the computing power service path. The computing node determination component decomposes the computing power service request instruction based on the computing power request instruction, obtains multiple sub-problems corresponding to the computing power service request, as well as the corresponding target algorithm type and computing mode, determines at least one computing task corresponding to each sub-problem, and determines at least one target computing power computing node for each sub-problem based on the computing power call relationship of computing power computing nodes. The association relationship establishment component sends computing power control instructions to the target computing power computing node, receives the computing results returned by the target computing power computing node, and establishes an association relationship between the target computing power computing node and the computing power service request. The above solution features on-demand, tiered scheduling capabilities that automatically match differentiated computing service paths for different user types, such as government agencies, enterprises, and research institutions (e.g., government needs may prioritize stability and security, while research needs focus more on high-performance computing). It can quickly identify the optimal service mode based on user identity, avoiding resource waste or insufficient computing power. Intelligent task decomposition and efficient scheduling allow the platform to automatically parse large computing demands into multiple parallel subtasks. Based on the characteristics of these subtasks (algorithm type, computing mode), it flexibly matches suitable computing nodes (e.g., GPU clusters are suitable for deep learning, while CPU clusters are suitable for traditional numerical computing), forming temporary computing alliances. Traceable end-to-end management ensures that each computing stage is controllable and traceable through the issuance and retrieval of computing power commands; binding computing results to the original requirements guarantees task integrity.

[0054] In summary, this embodiment can understand the personalized needs of different users (government / enterprise / research), dynamically organize distributed computing resources to complete tasks, and finally ensure that the service is auditable and optimizable through the association mechanism, thereby achieving refined operation of computing resources while ensuring efficiency.

[0055] Example 2: As Figure 2 As shown, based on Embodiment 1, the computing power request issuing component provided in this embodiment of the invention includes:

[0056] The type confirmation module is responsible for obtaining the computing power service request and the corresponding input method. The computing power service request triggers the user type identification program. The user type identification program determines the user type of the computing power service request based on the input method, and determines the performance range of the required computing power service path based on the user type.

[0057] The path searching module is responsible for constructing a computing power service path database corresponding to different user types, and the computing power service path database takes a performance interval as an index to determine a computing power service path corresponding to a user inputting a computing power service request;

[0058] The instruction conversion module is responsible for obtaining a computing power request instruction in response to the computing power service request, and the computing power request instruction contains a user type and a corresponding computing power service path.

[0059] The working principle and beneficial effects of the above technical solution are as follows: The type confirmation module of the embodiment obtains a computing power service request and a corresponding input mode, triggers a user type identification program through the computing power service request, determines a user type of the computing power service request according to the input mode, and determines a performance interval of a required computing power service path according to the user type; the path searching module constructs a computing power service path database corresponding to different user types, and the computing power service path database takes a performance interval as an index to determine a computing power service path corresponding to a user inputting a computing power service request; and the instruction conversion module obtains a computing power request instruction in response to the computing power service request, and the computing power request instruction contains a user type and a corresponding computing power service path. The type confirmation module of the above scheme identifies the user type of the computing power service request, and can accurately identify the user type through the input mode of the user. The path searching module matches the most suitable computing power service path for the user according to the user type and performance requirements by searching the database, and can ensure that each user can obtain computing power service meeting the requirements according to different performance intervals. The instruction conversion module generates a computing power request instruction containing a user type and a corresponding computing power service path in response to the computing power service request of the user after the computing power service path is determined, and ensures that the request of the user is accurately converted into a system executable instruction.

[0060] In summary, the embodiment provides customized computing power service paths according to the requirements of different users to improve user experience, accelerates the processing speed of computing power service requests through a pre-constructed database and a modular processing flow, allocates appropriate computing power resources for different types of users to avoid resource waste and improve resource utilization, and guarantees the stability and reliability of the system through modular design.

[0061] Embodiment 3: Based on embodiment 2, the path searching module provided by the embodiment of the application comprises:

[0062] The performance value definition submodule is responsible for establishing a computing power service path database containing a demand computing power computing node table, a response computing power computing node table, a computing power service path table, and a performance interval index; each demand computing power computing node and response computing power computing node is provided with a performance value;

[0063] The path end value confirmation submodule is responsible for, when the index of the performance interval is received, first querying the target demand computing node corresponding to the minimum performance value of the performance interval, taking the target demand computing node as the starting point of the computing power service path; and then querying the target response computing node corresponding to the maximum performance value of the performance interval, taking the target response computing node as the end point of the computing power service path.

[0064] The path output submodule is responsible for connecting the target demand computing node and the target response computing node to obtain the computing power service path, that is, determining the computing power service path corresponding to the user inputting the computing power service request.

[0065] The working principle and beneficial effects of the above technical solution are as follows: The performance value definition submodule of the embodiment establishes a computing power service path database including a demand computing node table, a response computing node table, a computing power service path table, and a performance interval index; each demand computing node and response computing node is provided with a performance value; when the path end value confirmation submodule receives the index of the performance interval, it first queries the target demand computing node corresponding to the minimum performance value of the performance interval, taking the target demand computing node as the starting point of the computing power service path; and then queries the target response computing node corresponding to the maximum performance value of the performance interval, taking the target response computing node as the end point of the computing power service path; the path output submodule connects the target demand computing node and the target response computing node to obtain the computing power service path, that is, determines the computing power service path corresponding to the user inputting the computing power service request. The performance value definition submodule of the above solution provides basic data support and performance parameters for computing power service path search, facilitates path search and end value confirmation according to performance values, and improves the accuracy and efficiency of computing power service path search. The path end value confirmation submodule quickly locates the appropriate starting and ending computing nodes according to the performance interval, ensures that the performance of the computing power service path meets the user's demand, reduces the search range of the path, and improves the efficiency of path search. The path output submodule finally determines the corresponding path of the user inputting the computing power service request, provides the user with the computing power service path information required by the user, and completes the entire process from performance value definition to path search.

[0066] In summary, the embodiment constructs a complete computing power service path search framework, including performance value definition, path end value confirmation, and path output; quickly and accurately locates the appropriate computing power service path according to the user's performance demand; improves the efficiency and accuracy of computing power service path search, meets the user's demand for high-performance computing power service; provides the user with a convenient and efficient computing power service path solution, and improves the user experience.

[0067] Embodiment 4: as Figure 3As shown, on the basis of Embodiment 1, the computing node determination component provided by the present embodiment comprises:

[0068] The task decomposition module is responsible for receiving the algorithm request instruction, calling the instruction decomposition program, decomposing the algorithm request instruction, obtaining the task required by the algorithm request instruction, and the task comprising a sub-problem, a target algorithm type and an algorithm mode.

[0069] The task establishment module is responsible for establishing a computing task corresponding to the sub-problem according to the matching relationship of the target algorithm type and the algorithm mode, and screening at least one computing task from the plurality of computing tasks.

[0070] The node matching module is responsible for calculating the algorithm power of each algorithm power computing node comprising a processing speed and a parallel processing capability, matching the computing demand in the sub-problem corresponding computing task with the algorithm power, obtaining the algorithm power computing node meeting the algorithm power, and determining the target algorithm power computing node processing the sub-problem by using the task scheduling strategy.

[0071] The working principle and beneficial effects of the above technical solution are as follows: the task decomposition module of the present embodiment receives the algorithm request instruction, calls the instruction decomposition program, decomposes the algorithm request instruction, obtains the task required by the algorithm request instruction, and the task comprising a sub-problem, a target algorithm type and an algorithm mode; the task establishment module establishes a computing task corresponding to the sub-problem according to the matching relationship of the target algorithm type and the algorithm mode, and screens at least one computing task from the plurality of computing tasks; the node matching module calculates the algorithm power of each algorithm power computing node comprising a processing speed and a parallel processing capability, matches the computing demand in the sub-problem corresponding computing task with the algorithm power, obtains the algorithm power computing node meeting the algorithm power, and determines the target algorithm power computing node processing the sub-problem by using the task scheduling strategy. The task decomposition module of the above solution helps to clarify the computing demand of each sub-problem, including the required algorithm type and mode. The task establishment module creates a computing task corresponding to each sub-problem, and screens out a suitable task from a plurality of possible computing tasks to meet the computing demand of the specific sub-problem. The node matching module matches the processing speed and parallel processing capability of each computing node with the computing demand of the corresponding computing task; the task scheduling strategy is used to ensure that each sub-problem is allocated to the most suitable algorithm power computing node for execution.

[0072] In summary, the present embodiment accurately matches the computing demand and the computing resource, avoids resource waste, and ensures efficient use of computing resources; by decomposing complex computing tasks and reasonably allocating them to different algorithm power computing nodes, the speed of the entire computing process is accelerated; the task allocation can be adjusted according to actual conditions, and suitable computing resources can be flexibly selected for different sub-problems and algorithm types; by reasonably allocating tasks, the problem of some computing nodes being overloaded while other nodes being idle is avoided, and load balancing is achieved.

[0073] In the embodiment 5, based on the embodiment 4, the task decomposition module comprises:

[0074] The program content setting submodule is responsible for taking the computing power request instruction as a trigger event of the instruction decomposition program, and is used for task decomposition of the computing power request instruction. The running instruction decomposition program is obtained by compiling and executing the first code of the instruction decomposition program. The first code comprises calling code of the candidate running code in the second code. The program code of the instruction decomposition program comprises the first code and the second code. The first part of the code is the code for displaying the tasks in the computing power request instruction. The second code is the code for classifying and integrating the task attributes.

[0075] The decomposition list generation submodule is responsible for executing the first code to display the tasks in the computing power request instruction, decomposing the tasks into sub-problems according to the time nodes, assigning the tasks according to the sequence of the time nodes, and generating a task decomposition list.

[0076] The task classification and integration submodule is responsible for loading the task attributes in the task decomposition list, calling the second code to classify and integrate the task attributes, generating a task decomposition result set, and obtaining the sub-problems, the target algorithm type and the algorithm mode under different time nodes.

[0077] The working principle and beneficial effects of the above technical solution are as follows: the program content setting submodule of the embodiment takes the computing power request instruction as a trigger event of the instruction decomposition program, and is used for task decomposition of the computing power request instruction. The running instruction decomposition program is obtained by compiling and executing the first code of the instruction decomposition program. The first code comprises calling code of the candidate running code in the second code. The program code of the instruction decomposition program comprises the first code and the second code. The first part of the code is the code for displaying the tasks in the computing power request instruction. The second code is the code for classifying and integrating the task attributes. The decomposition list generation submodule executes the first code to display the tasks in the computing power request instruction, decomposes the tasks into sub-problems according to the time nodes, assigns the tasks according to the sequence of the time nodes, and generates a task decomposition list. The task classification and integration submodule loads the task attributes in the task decomposition list, calls the second code to classify and integrate the task attributes, generates a task decomposition result set, and obtains the sub-problems, the target algorithm type and the algorithm mode under different time nodes. The program content setting submodule is responsible for starting the entire task decomposition process when the computing power request instruction arrives. The decomposition list generation submodule decomposes the tasks into a list and lists each task. The task classification and integration submodule loads the task decomposition list and calls the second code to classify and integrate the task attributes.

[0078] Embodiment 6: Based on Embodiment 4, the node matching module provided by the embodiment of the application comprises:

[0079] The index data calculation submodule is responsible for acquiring the computing power resource index data of all computing power calculation nodes, including processor use efficiency, memory usage rate, load rate and the like.

[0080] The prediction data calculation submodule is responsible for constructing a computing power calculation correlation graph according to the computing power resource index data, and predicting computing power demand prediction data according to the computing demand in the calculation task corresponding to the sub-problems.

[0081] The resource score calculation submodule is responsible for obtaining the computing power resource score of the computing power calculation node based on the computing power calculation correlation graph and the computing power demand prediction data, and obtaining the computing power calculation node that meets the computing power.

[0082] The working principle and beneficial effects of the above technical solution are as follows: the index data calculation submodule of the embodiment acquires the computing power resource index data of all computing power calculation nodes, including processor use efficiency, memory usage rate, load rate and the like; the prediction data calculation submodule constructs a computing power calculation correlation graph according to the computing power resource index data, and predicts computing power demand prediction data according to the computing demand in the calculation task corresponding to the sub-problems; and the resource score calculation submodule obtains the computing power resource score of the computing power calculation node based on the computing power calculation correlation graph and the computing power demand prediction data, and obtains the computing power calculation node that meets the computing power. The index data calculation submodule of the above solution establishes an accurate mapping system from physical resources to quantifiable parameters; the prediction data calculation submodule upgrades the static resource evaluation to a dynamic demand prediction system; and the resource score calculation submodule has a multi-dimensional evaluation system with dynamic weight adjustment capability.

[0083] Embodiment 7: As shown in Figure 4 Based on Embodiments 1-6, the computing power service method provided by the embodiment of the application comprises the following steps:

[0084] S100: Acquire the user types of government, enterprise and scientific research institution and the like of the computing power service, and assign different computing power service paths to different user types; when the user issues a computing power service demand, obtain the corresponding computing power service path according to the user type, and issue a computing power request instruction according to the computing power service path;

[0085] S200: According to the computing power request instruction, decompose the computing power service request instruction to obtain a plurality of sub-problems corresponding to the computing power service request, and the corresponding target algorithm type and calculation mode, determine at least one calculation task corresponding to each sub-problem, and determine at least one target computing power calculation node for each sub-problem according to the computing power calculation node computing power calling relationship;

[0086] S300: send a computing power control instruction to the target computing power computing node, receive a computing result returned by the target computing power computing node, and establish an association relationship between the target computing power computing node and the computing power service demand.

[0087] The working principle and beneficial effects of the technical solution are as follows: the embodiment first acquires user types of government, enterprise, and scientific research institution of the computing power service, and gives different computing power service paths to different user types; when a user issues a computing power service demand, a corresponding computing power service path is obtained according to the user type, and a computing power request instruction is issued according to the computing power service path; secondly, the computing power request instruction is decomposed according to the computing power request instruction, a plurality of sub-problems corresponding to the computing power service request, and corresponding target algorithm types and computing modes are obtained, at least one computing task corresponding to each sub-problem is determined, at least one target computing power computing node is determined for each sub-problem according to the computing power computing node computing power calling relationship; finally, a computing power control instruction is sent to the target computing power computing node, a computing result returned by the target computing power computing node is received, and an association relationship between the target computing power computing node and the computing power service demand is established. The above scheme can not only understand the individualized demands of different users (government / enterprise / scientific research), but also dynamically organize distributed computing resources to complete tasks, and finally ensure that the service is auditable and optimal through the association mechanism, ensuring efficiency while realizing fine operation of computing power resources.

[0088] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of equivalent technology of the present application, the present application also intends to include these modifications and variations.

Claims

1. A computing power service platform, characterized in that, Include: The computing power request issuing component is responsible for obtaining the user types of government, enterprises and research institutions that provide computing power services, and assigning different computing power service paths to different user types; when a user issues a computing power service request, the component obtains the corresponding computing power service path according to the user type, and issues a computing power request instruction according to the computing power service path. The component that issues computing power requests includes: The type confirmation module is responsible for obtaining the computing power service request and the corresponding input method. The computing power service request triggers the user type identification program. The user type identification program determines the user type of the computing power service request based on the input method, and determines the performance range of the required computing power service path based on the user type. The path search module is responsible for building a computing power service path database corresponding to different user types. The computing power service path database uses performance range as an index to determine the computing power service path corresponding to the user who inputs a computing power service request. The instruction conversion module is responsible for responding to computing power service requests and obtaining computing power request instructions, which include user type and corresponding computing power service path; The path search module includes: The performance value definition submodule is responsible for establishing the computing power service path database; each demand computing power computing node and response computing power computing node is assigned a performance value. The path endpoint confirmation submodule is responsible for, upon receiving the index of the performance range, first querying the target required computing power computing node corresponding to the minimum performance value of the performance range, and using the target required computing power computing node as the starting point of the computing power service path; then querying the target response computing power computing node corresponding to the maximum performance value of the performance range, and using the target response computing power computing node as the ending point of the computing power service path. The path output submodule is responsible for connecting the target computing power demand node and the target computing power response node to obtain the computing power service path, that is, determining the computing power service path corresponding to the user who inputs the computing power service request; The computing node determination component is responsible for decomposing the computing power service request instruction according to the computing power request instruction, and obtaining multiple sub-problems corresponding to the computing power service request, as well as the corresponding target algorithm type and computing mode; The association establishment component is responsible for sending computing power control commands to the target computing power node, receiving the computing results returned by the target computing power node, and establishing the association between the target computing power node and the computing power service requirement.

2. The computing power service platform as described in claim 1, characterized in that, The computing power service path database includes a demand computing power computing node table, a response computing power computing node table, a computing power service path table, and a performance range index.

3. The computing power service platform as described in claim 1, characterized in that, The compute node determination component includes: The task decomposition module is responsible for receiving computing power request instructions and calling the instruction decomposition program to decompose the computing power request instructions into tasks, which include sub-problems, target algorithm types and algorithm modes. The task creation module is responsible for creating computational tasks corresponding to sub-problems based on the matching relationship between the target algorithm type and algorithm mode, and selecting at least one computational task from multiple computational tasks. The node matching module is responsible for calculating the computing power of each computing node, including processing speed and parallel processing capabilities. It matches the computing power with the computing requirements in the corresponding computing tasks of sub-problems to obtain computing nodes that meet the computing power requirements. A task scheduling strategy is adopted to determine the target computing nodes for processing sub-problems.

4. The computing power service platform as described in claim 3, characterized in that, The task decomposition module includes: The program content setting submodule is responsible for using the computing power request instruction as the trigger event for the instruction decomposition program. It is used to decompose the computing power request instruction into tasks. The instruction decomposition program is obtained by compiling and executing the first code of the instruction decomposition program. The task breakdown list generation submodule is responsible for executing the first code, displaying the tasks in the computing power request instruction, breaking down the computing power service request into sub-problems according to the time nodes, assigning tasks in the order of the time nodes, and generating a task breakdown list. The task classification and integration submodule is responsible for loading the task attributes in the task decomposition list, calling the second code to classify and integrate the task attributes, generating a task decomposition result set, and obtaining the sub-problems, target algorithm types and algorithm modes at different time points.

5. The computing power service platform as described in claim 4, characterized in that, The first code includes the calling code of the candidate running code in the second code. The program code of the instruction decomposition program includes the first code and the second code. The first part of the code is the code that compiles the instructions to display the tasks in the computing power request instructions. The second code is the code that classifies, integrates and executes the task attributes.

6. The computing power service platform as described in claim 3, characterized in that, The node matching module includes: The indicator data calculation submodule is responsible for obtaining the computing resource indicator data of all computing nodes, including processor utilization efficiency, memory utilization rate and load rate. The prediction data calculation submodule is responsible for constructing a computing power calculation correlation diagram based on computing power resource index data, and predicting computing power demand prediction data based on the computing needs in the corresponding computing tasks of sub-problems. The resource scoring calculation submodule is responsible for obtaining the computing resource score of computing nodes based on the computing power computing relationship graph and computing power demand prediction data, and thus obtaining computing nodes that meet the computing power requirements.

7. The computing power service platform as described in claim 1, characterized in that, The computing node determination component determines at least one computing task corresponding to each sub-problem, and determines at least one target computing node for each sub-problem based on the computing power call relationship of the computing power computing nodes.

8. A computing power service method, characterized in that, Includes the following steps: The system identifies the types of government, enterprise, and research institution users who require computing power services and assigns different computing power service paths to different user types. When a user issues a computing power service request, the system obtains the corresponding computing power service path based on the user type and issues a computing power request instruction based on the computing power service path. The type confirmation module is responsible for obtaining the computing power service request and the corresponding input method. The computing power service request triggers the user type identification program. The user type identification program determines the user type of the computing power service request based on the input method, and determines the performance range of the required computing power service path based on the user type. The path search module is responsible for building a computing power service path database corresponding to different user types. The computing power service path database uses performance range as an index to determine the computing power service path corresponding to the user who inputs a computing power service request. The instruction conversion module is responsible for responding to computing power service requests and obtaining computing power request instructions, which include user type and corresponding computing power service path; The path search module includes: The performance value definition submodule is responsible for establishing the computing power service path database; each demand computing power computing node and response computing power computing node is assigned a performance value. When the path endpoint confirmation submodule receives the index of the performance range, it first queries the target required computing power computing node corresponding to the minimum performance value of the performance range, and uses the target required computing power computing node as the starting point of the computing power service path; then it queries the target response computing power computing node corresponding to the maximum performance value of the performance range, and uses the target response computing power computing node as the ending point of the computing power service path. The path output submodule is responsible for connecting the target computing power demand node and the target computing power response node to obtain the computing power service path, that is, to determine the computing power service path corresponding to the user who inputs the computing power service request; Based on the computing power request instruction, the computing power service request instruction is decomposed to obtain multiple sub-problems corresponding to the computing power service request, as well as the corresponding target algorithm type and computing mode. At least one computing task is determined for each sub-problem. Based on the computing power call relationship of the computing power computing nodes, at least one target computing power computing node is determined for each sub-problem. Send computing power control commands to the target computing power node, receive the computing results returned by the target computing power node, and establish the association between the target computing power node and the computing power service demand.

Images