Heterogeneous task scheduling method and system oriented to intelligent computing center and electronic equipment

Abstract

The invention provides an intelligent computing center-oriented heterogeneous task scheduling method and system and electronic equipment. The method comprises the following steps of: acquiring computing power resource information of each computing power node in an intelligent computing center; determining task information and computing power resource requirements of each heterogeneous task; according to a preset scheduling rule, on the basis of the computing power resource information, the task information and the computing power resource demand, generating a scheduling strategy, and on the basis of the scheduling strategy, scheduling each heterogeneous task to each corresponding target computing node; wherein the preset scheduling rule comprises at least one of the following items: based on a priority allocation algorithm, performing computing power resource allocation on each heterogeneous task; based on a heuristic algorithm, carrying out computing power resource allocation on each heterogeneous task; and according to the preset load threshold value of each computing power node, carrying out computing power resource allocation on each heterogeneous task. According to the method, a proper scheduling mode can be flexibly selected for task scheduling, the computing power resource utilization rate is optimized, the service quality of large-scale heterogeneous tasks can be guaranteed, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a heterogeneous task scheduling method, system, and electronic equipment for an intelligent computing center. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, AI model parameters are growing exponentially, requiring vast amounts of computing resources to train complex models with vast parameters and support a vast array of diverse model inference tasks. However, intelligent computing centers have limited computing resources, and how to rationally allocate computing resources and improve resource utilization is a key issue that needs to be addressed. Existing heterogeneous task scheduling solutions for intelligent computing centers typically utilize a single scheduling method, resulting in low resource utilization and difficulty ensuring the quality of service required for large-scale heterogeneous tasks, impacting user experience. Summary of the Invention

[0003] The present invention provides a heterogeneous task scheduling method, system and electronic device for intelligent computing centers, which are used to solve the defects of the existing technology that heterogeneous task scheduling solutions for intelligent computing centers usually only use a single scheduling method to schedule tasks, have low resource utilization, and are difficult to guarantee the service quality requirements of large-scale heterogeneous tasks.

[0004] In a first aspect, the present invention provides a heterogeneous task scheduling method for an intelligent computing center, comprising: Obtain computing power resource information of each computing power node in the intelligent computing center; Determine the task information and computing resource requirements of each heterogeneous task to be processed; Generate a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to corresponding target computing nodes based on the scheduling strategy; The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

[0005] In some embodiments, the method further comprises: After the heterogeneous tasks are executed, feedback information is obtained; Based on the feedback information, the preset scheduling rule is optimized, and based on the optimized preset scheduling rule, the scheduling strategy is optimized.

[0006] In some embodiments, the optimizing the preset scheduling rule based on the feedback information, and optimizing the scheduling strategy based on the optimized preset scheduling rule, includes: Based on the feedback information, evaluating the execution effects of the heterogeneous tasks to obtain an evaluation result; Based on the evaluation result, the preset scheduling rule is optimized to obtain an optimized preset scheduling rule; Obtain updated computing resource information of each computing node; According to the optimized preset scheduling rules, based on the updated computing resource information, the task information and the computing resource requirements, the scheduling strategy is optimized to obtain an optimized scheduling strategy; Based on the optimized scheduling strategy, the heterogeneous tasks are scheduled.

[0007] In some embodiments, the optimizing the preset scheduling rule includes at least one of the following: Optimizing the priority allocation algorithm; Optimizing the heuristic algorithm; Modify the preset load threshold of each computing power node.

[0008] In some embodiments, generating a scheduling strategy according to a preset scheduling rule based on the computing resource information, task information, and computing resource requirements includes: Screening the computing power nodes to obtain multiple candidate computing power nodes; Matching the computing power resource information of the multiple candidate computing power nodes with the task information and computing power resource requirements of the heterogeneous tasks to obtain a matching result; Based on the preset scheduling rules and the matching results, a scheduling strategy for each of the heterogeneous tasks is generated.

[0009] In some embodiments, the heuristic algorithm-based allocation of computing resources to heterogeneous tasks includes: Construct an objective function with the goal of minimizing the total task execution time and / or maximizing the total resource utilization; Based on a heuristic algorithm, the objective function is solved to obtain a scheduling strategy for each heterogeneous task.

[0010] In some embodiments, the computing resource information includes the computing resource type, the amount of used computing resources, and the amount of available computing resources; the task information includes the task type and task indicators; and the computing resource demand includes the computing resource demand and the computing resource type demand.

[0011] In a second aspect, the present invention further provides a heterogeneous task scheduling system for an intelligent computing center, comprising: Resource perception module, used to obtain computing resource information of each computing node in the intelligent computing center; Resource requirement assessment module, used to determine the task information and computing resource requirements of each heterogeneous task to be processed; A scheduling module is configured to generate a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to corresponding target computing nodes based on the scheduling strategy; The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

[0012] In some embodiments, the system further comprises: A feedback module, configured to obtain feedback information after the execution of each heterogeneous task; The optimization module is used to optimize the preset scheduling rules based on the feedback information, and optimize the scheduling strategy based on the optimized preset scheduling rules.

[0013] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a heterogeneous task scheduling method for an intelligent computing center as described in any one of the above.

[0014] The heterogeneous task scheduling method, system and electronic device for the intelligent computing center provided by the present invention obtain the computing power resource information of each computing power node in the intelligent computing center; determine the task information and computing power resource requirements of each heterogeneous task to be processed; according to preset scheduling rules, flexibly select appropriate scheduling methods from priority scheduling methods, heuristic algorithm-based scheduling methods and threshold scheduling methods for combination; on this basis, combine the computing power resource information, task information and computing power resource requirements to generate a scheduling strategy; based on the scheduling strategy, each heterogeneous task is scheduled to the corresponding target computing node, thereby optimizing the utilization of node resources, ensuring the service quality of large-scale heterogeneous tasks in complex application scenarios, and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 This is one of the flow charts of the heterogeneous task scheduling method for the intelligent computing center provided by an embodiment of the present invention.

[0017] Figure 2 This is the second flow chart of the heterogeneous task scheduling method for the intelligent computing center provided by an embodiment of the present invention.

[0018] Figure 3 It is a structural diagram of a heterogeneous task scheduling system for intelligent computing centers provided by an embodiment of the present invention.

[0019] Figure 4 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0021] The terms "first," "second," and the like, as used herein, are used to distinguish similar objects, and are not intended to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable, where appropriate, so that embodiments of the present invention can be implemented in an order other than that illustrated or described herein. Furthermore, the objects distinguished by "first" and "second" generally refer to a class of objects and do not limit the number of objects. For example, the first object can be one or more. Furthermore, the term "and / or" as used herein refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0022] Figure 1 This is a flow chart of a heterogeneous task scheduling method for an intelligent computing center provided by an embodiment of the present invention. Figure 1 As shown, a heterogeneous task scheduling method for intelligent computing centers is provided, including the following steps: step 110, step 120, and step 130. The steps of the method flow are only a possible implementation of the present invention.

[0023] Step 110: Obtain computing resource information of each computing node in the intelligent computing center; Step 120: Determine the task information and computing resource requirements of each heterogeneous task to be processed; Step 130: Generate a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to the corresponding target computing nodes based on the scheduling strategy. The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

[0024] Among them, the intelligent computing center is a data center mainly used for AI computing tasks. It integrates advanced computing technology, storage technology and network technology to provide computing power support for efficient operation of AI algorithms and models. The intelligent computing center uses specialized AI computing power hardware, such as graphics processing units GPU, neural network processors NPU, tensor processing units TPU, etc. These chips have powerful parallel computing capabilities and can quickly process complex matrix operations and deep learning tasks. The intelligent computing center usually adopts heterogeneous computing architecture, such as CPU+GPU+NPU.

[0025] A computing node in an intelligent computing center refers to a single computing unit or a collection of computing units that can provide computing power. A computing node is the basic unit that actually performs computing tasks in the intelligent computing center, providing the computing resources required for various AI algorithms and applications. Computing nodes are typically composed of high-performance computing devices such as CPUs, GPUs, and NPUs.

[0026] Among them, heterogeneous tasks refer to various types of computing tasks. These tasks cannot be efficiently completed by a single type of computing power node. They need to combine the characteristics of different computing power nodes, such as the versatility and flexibility of the CPU, the parallel computing capabilities of the GPU, and the AI computing optimization of the NPU, to jointly achieve the task goals.

[0027] Optionally, heterogeneous tasks can be AI training tasks (such as deep learning model training, reinforcement learning model training), AI reasoning tasks (such as image recognition, natural language processing), or data processing and analysis tasks (such as big data preprocessing, data analysis and mining).

[0028] Optionally, the task information includes but is not limited to: task ID, task name, task type, task priority, task indicators, etc.

[0029] In some embodiments, computing power resource information includes computing power resource type, used computing power resource amount and available computing power resource amount; task information includes task type and task indicators; computing power resource demand includes computing power resource amount demand and computing power resource type demand.

[0030] Optionally, computing power resource types include but are not limited to: basic computing power, intelligent computing power, and super computing power.

[0031] Among them, the basic computing power is mainly based on the output of the general computing unit CPU, providing computing power support for common computing tasks, and is the most basic form of computing resources; the basic computing power is suitable for various types of computing tasks, such as conventional data processing, simple application running, etc., with strong versatility, strong flexibility and good stability.

[0032] Among them, intelligent computing power is based on customized chips for artificial intelligence such as GPU, NPU, and TPU, providing powerful computing support for AI applications; super computing power mainly relies on high-performance computing clusters or dedicated supercomputers, combined with advanced algorithms and software technologies, to provide extremely powerful computing capabilities for scientific research, complex engineering problems, and large-scale simulations.

[0033] Among them, task types include but are not limited to: model training, model reasoning, data preprocessing, data mining, etc.

[0034] Optionally, the task type may be computationally intensive, memory-intensive, or the like.

[0035] Among them, task indicators include but are not limited to: performance indicators (such as execution time, resource utilization, etc.), quality indicators (such as accuracy), cost indicators, etc.

[0036] Among them, common types of heuristic algorithms include traditional heuristic algorithms, such as greedy algorithms, hill climbing algorithms, etc., as well as meta-heuristic algorithms, such as genetic algorithms, ant colony algorithms, particle swarm optimization algorithms, etc. In addition, there are super-heuristic algorithms, which are composed of a series of heuristic algorithms and are more intelligent.

[0037] It's important to note that allocating computing resources to heterogeneous tasks based on heuristic algorithms requires precise modeling and problem definition, including analyzing task characteristics, describing computing resources, and determining the objective function. An appropriate heuristic algorithm is then selected based on the problem and a scheduling solution is initialized. This is followed by iterative optimization through evaluation, search adjustments, and termination judgment. Finally, a scheduling solution is output and verified for effectiveness, ensuring efficient task completion and optimal resource utilization, providing a viable path for allocating computing resources to complex, heterogeneous tasks.

[0038] Optionally, the scheduling strategy is: first, based on the priority allocation algorithm, high, medium and low priority tasks are allocated in sequence. During the allocation process, a heuristic algorithm is used to allocate tasks. The heuristic algorithm selects the best fit packing strategy to prioritize filling node resources. In the process of allocating tasks one by one, combined with the threshold scheduling method, when the load level of a node exceeds a threshold, the node is deleted from the list of schedulable nodes to keep the node load level within the set threshold.

[0039] In an embodiment of the present invention, computing power resource information of each computing power node in the intelligent computing center is obtained; task information and computing power resource requirements of each heterogeneous task to be processed are determined; according to preset scheduling rules, appropriate scheduling methods are flexibly selected from priority scheduling methods, heuristic algorithm-based scheduling methods and threshold scheduling methods for combination. On this basis, a scheduling strategy is generated by combining computing power resource information, task information and computing power resource requirements. Based on the scheduling strategy, each heterogeneous task is scheduled to the corresponding target computing node, thereby optimizing the utilization of node resources, ensuring the service quality of large-scale heterogeneous tasks in complex application scenarios, and improving user experience.

[0040] In some embodiments, the above method further comprises: After each heterogeneous task is executed, feedback information is obtained; Based on the feedback information, the preset scheduling rules are optimized, and based on the optimized preset scheduling rules, the scheduling strategy is optimized.

[0041] The feedback information includes at least: calculation accuracy, calculation efficiency, resource utilization, task execution status, error or warning information, result stability, etc.

[0042] It is understandable that optimizing preset scheduling rules based on feedback information and improving scheduling strategies accordingly can comprehensively improve computing resource utilization, reduce waste and avoid node overload, improve task execution efficiency, enhance system stability and reliability, reduce node failures, reduce the risk of task execution interruption, and reduce computing costs.

[0043] In some embodiments, optimizing the preset scheduling rules includes at least one of the following: Optimize the priority allocation algorithm; Optimize the heuristic algorithm; Modify the preset load threshold of each computing power node.

[0044] Optionally, during task execution, key characteristics of the task, such as computational complexity and data volume, are continuously monitored. For example, for a data processing task, the amount of data required may increase as data continues to be input. If the data volume exceeds a certain threshold, the priority of the task is dynamically increased to ensure that it can be completed within a reasonable time.

[0045] Optionally, the task priority is adjusted based on the usage of computing resources (such as CPU, GPU, etc.).

[0046] Optionally, a multi-dimensional priority evaluation is performed on each heterogeneous task to adjust the task priority.

[0047] Optionally, you can customize priority rules based on user needs. For example, you can specify that certain types of tasks (such as those involving critical business logic) have higher priorities.

[0048] Alternatively, different heuristic algorithms can be combined to leverage their respective strengths. For example, the global search capabilities of a genetic algorithm can be combined with the rapid convergence of a particle swarm optimization algorithm. Initially, the genetic algorithm can be used to search the solution space extensively to identify promising solutions. Based on these solutions, the particle swarm optimization algorithm can then be used to perform a local, refined search to improve the quality of the solutions.

[0049] Optionally, the heuristic algorithm can be adaptively switched based on the different stages of the problem or the state of the solution. For example, in the early stages of the problem, when the solution space is large and promising areas need to be quickly located, a simulated annealing algorithm can be used for a rough search. When the optimal solution is near, a greedy algorithm can be switched for precise adjustments to accelerate convergence.

[0050] Optionally, the parameters of the heuristic algorithm are dynamically adjusted according to the characteristics of the problem and feedback information during the iteration process; or some optimization methods are used to determine the optimal parameters of the heuristic algorithm.

[0051] Optionally, the monitoring software installed on the computing power node can obtain the node's load information in real time, and dynamically update the node's preset load threshold based on the statistical results of the node's historical load data.

[0052] Figure 2 The second flow chart of the heterogeneous task scheduling method for intelligent computing center provided by the embodiment of the present invention. Figure 2 As shown, in some embodiments, based on the feedback information, the preset scheduling rules are optimized, and based on the optimized preset scheduling rules, the scheduling strategy is optimized, including: Based on the feedback information, the execution effect of each heterogeneous task is evaluated to obtain the evaluation results; Based on the evaluation results, the preset scheduling rules are optimized to obtain optimized preset scheduling rules; Obtain updated computing resource information of each computing node; According to the optimized preset scheduling rules, based on the updated computing resource information, as well as the task information and computing resource requirements, the scheduling strategy is optimized to obtain the optimized scheduling strategy; Based on the optimized scheduling strategy, various heterogeneous tasks are scheduled.

[0053] Optionally, evaluation indicators such as task completion time, resource utilization, task quality, etc. are determined, and based on the feedback information, the execution effect of each heterogeneous task is evaluated according to the evaluation indicators.

[0054] For example, compare the actual completion time of a task with the preset expected completion time. If the actual time is much longer than expected, it means that the task encountered problems during execution, such as insufficient resources or inefficient algorithms.

[0055] Optionally, each evaluation indicator is calculated and quantified based on the collected feedback information, and the execution effect of the task is divided into different categories, such as excellent, good, average, poor, etc., according to the indicator value and preset standards.

[0056] Optionally, when multiple heterogeneous tasks are executed simultaneously, evaluate overall resource utilization. Analyze whether there is resource contention or idleness between tasks, and how to optimize resource allocation to improve overall resource utilization.

[0057] In an embodiment of the present invention, by adopting a multi-strategy joint scheduling method to optimize the scheduling strategy for allocating large-scale heterogeneous tasks to nodes in an intelligent computing center, the overall resource utilization of the intelligent computing center can be effectively improved while meeting the task quality requirements. Specifically, this method combines the resource status and characteristics of the actual operation of the intelligent computing center to simultaneously consider and evaluate multi-dimensional resources; by prioritizing tasks, a combination of priority strategy, threshold strategy, and heuristic algorithm strategy is adopted to optimize node resource utilization while ensuring the task service quality requirements.

[0058] In some embodiments, according to preset scheduling rules, based on computing resource information, task information, and computing resource requirements, a scheduling strategy is generated, including: Screen each computing power node to obtain multiple candidate computing power nodes; Match the computing resource information of multiple candidate computing nodes with the task information and computing resource requirements of each heterogeneous task to obtain a matching result; Based on the preset scheduling rules and matching results, the scheduling strategy for each heterogeneous task is generated.

[0059] Optionally, the computing power nodes are screened based on the computing power, storage capacity, and network bandwidth required for the task, and computing power nodes that meet the task requirements are screened out.

[0060] Optionally, each computing node is screened based on its current load, stability, and reliability, with priority given to computing nodes with lower load, stable operation, and low failure rate.

[0061] Optionally, for computing-intensive tasks, such as scientific computing, encryption and decryption, etc., tasks are assigned to nodes with strong computing capabilities; for data-intensive tasks, such as data mining, image and video processing, etc., nodes with large storage capacity and high read and write speeds are matched; for communication-intensive tasks, such as service calls and message passing in distributed systems, nodes with high network bandwidth and low latency are selected.

[0062] In some embodiments, computing resources are allocated to heterogeneous tasks based on a heuristic algorithm, including: Construct an objective function with the goal of minimizing the total task execution time and / or maximizing the total resource utilization; Based on the heuristic algorithm, the objective function is solved to obtain the scheduling strategy for each heterogeneous task.

[0063] Optionally, determine the key characteristics of each heterogeneous task, such as computational complexity, data volume, task priority, etc.

[0064] Optionally, model the available computing resources and identify key performance indicators for each computing resource. For example, for a CPU, you can use clock speed and number of cores to measure its computing power; for a GPU, you can consider the number of stream processors and memory bandwidth.

[0065] Optionally, choose an appropriate heuristic algorithm based on the characteristics and scale of the problem. For example, for small-scale problems, use a simple greedy algorithm; for complex, large-scale resource allocation problems, use a more sophisticated meta-heuristic algorithm such as a genetic algorithm or a particle swarm optimization algorithm.

[0066] It can be understood that by constructing an objective function with the goal of minimizing the total task execution time and / or maximizing the total resource utilization, and solving the objective function based on a heuristic algorithm, the scheduling strategy for each heterogeneous task is obtained. This is simple and easy, avoids a comprehensive search of the solution space, and can significantly save computing time and space resources when dealing with large-scale problems, improve computing efficiency, and has strong flexibility.

[0067] The following describes a heterogeneous task scheduling system for an intelligent computing center provided by an embodiment of the present invention. The heterogeneous task scheduling system for an intelligent computing center described below and the heterogeneous task scheduling method for an intelligent computing center described above can be referenced to each other.

[0068] Figure 3 A schematic diagram of the structure of a heterogeneous task scheduling system for an intelligent computing center provided by an embodiment of the present invention is shown in FIG. Figure 3 As shown, the heterogeneous task scheduling system 300 for intelligent computing centers includes: Resource sensing module 310, used to obtain computing resource information of each computing node in the intelligent computing center; Resource requirement assessment module 320, used to determine the task information and computing resource requirements of each heterogeneous task to be processed; The scheduling module 330 is used to generate a scheduling policy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to the corresponding target computing nodes based on the scheduling policy; The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

[0069] Among them, the resource perception module is used to perceive each node of the intelligent computing center resource information.

[0070] The resource sensing module includes a resource type sensing unit, which is used to obtain the resource type and total size of the node. T Indicates that the resource size is represented by a vector indicates; among them Representation node n Type of resources owned t In the intelligent computing center, node resource types mainly include memory, CPU cores, GPU memory, and computing resources.

[0071] Among them, the resource demand assessment module is used to evaluate the set of heterogeneous tasks requested to run in the intelligent computing center; the resource demand assessment module includes an indicator demand perception unit and a resource analysis unit.

[0072] Among them, the indicator requirement perception unit is used to obtain the indicator requirements of heterogeneous tasks, such as running completion time and AI task reasoning time.

[0073] The resource analysis unit is used to calculate the resource type and size required for the task based on the indicator requirements. The task set requested to run is defined as , each task s The requested resource size is .

[0074] Optionally, the above system further includes: Feedback module, used to obtain feedback information after the execution of various heterogeneous tasks; The optimization module is used to optimize the preset scheduling rules based on the feedback information, and optimize the scheduling strategy based on the optimized preset scheduling rules.

[0075] Optionally, based on the feedback information, the preset scheduling rules are optimized, and based on the optimized preset scheduling rules, the scheduling strategy is optimized, including: Based on the feedback information, the execution effect of each heterogeneous task is evaluated to obtain the evaluation results; Based on the evaluation results, the preset scheduling rules are optimized to obtain optimized preset scheduling rules; Obtain updated computing resource information of each computing node; According to the optimized preset scheduling rules, based on the updated computing resource information, as well as the task information and computing resource requirements, the scheduling strategy is optimized to obtain the optimized scheduling strategy; Based on the optimized scheduling strategy, various heterogeneous tasks are scheduled.

[0076] Optionally, optimizing the preset scheduling rules includes at least one of the following: Optimize the priority allocation algorithm; Optimize the heuristic algorithm; Modify the preset load threshold of each computing power node.

[0077] Optionally, according to preset scheduling rules, based on computing resource information, task information, and computing resource requirements, a scheduling strategy is generated, including: Screen each computing power node to obtain multiple candidate computing power nodes; Match the computing resource information of multiple candidate computing nodes with the task information and computing resource requirements of each heterogeneous task to obtain a matching result; Based on the preset scheduling rules and matching results, the scheduling strategy for each heterogeneous task is generated.

[0078] Optionally, computing resources are allocated to heterogeneous tasks based on a heuristic algorithm, including: Construct an objective function with the goal of minimizing the total task execution time and / or maximizing the total resource utilization; Based on the heuristic algorithm, the objective function is solved to obtain the scheduling strategy for each heterogeneous task.

[0079] Optionally, computing power resource information includes computing power resource type, used computing power resource amount and available computing power resource amount; task information includes task type and task indicators; computing power resource demand includes computing power resource amount demand and computing power resource type demand.

[0080] It should be noted here that the heterogeneous task scheduling system for intelligent computing centers provided by the embodiment of the present invention can implement all the method steps implemented by the above-mentioned heterogeneous task scheduling method embodiment for intelligent computing centers, and can achieve the same technical effects. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0081] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communications bus 440. The processor 410 may call logic instructions in the memory 430 to execute a heterogeneous task scheduling method for an intelligent computing center, the method comprising: obtaining computing resource information of each computing node in the intelligent computing center; determining task information and computing resource requirements of each heterogeneous task to be processed; generating a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and scheduling each heterogeneous task to a corresponding target computing node based on the scheduling strategy; wherein the preset scheduling rules include at least one of the following: allocating computing resources to each heterogeneous task based on a priority allocation algorithm; allocating computing resources to each heterogeneous task based on a heuristic algorithm; and allocating computing resources to each heterogeneous task based on a preset load threshold of each computing node.

[0082] Furthermore, the logic instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0084] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion 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, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A heterogeneous task scheduling method for intelligent computing centers, characterized in that: include: Obtain computing power resource information of each computing power node in the intelligent computing center; Determine the task information and computing resource requirements of each heterogeneous task to be processed; Generate a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to corresponding target computing nodes based on the scheduling strategy; The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

2. The heterogeneous task scheduling method for intelligent computing centers according to claim 1 is characterized in that: The method further comprises: After the heterogeneous tasks are executed, feedback information is obtained; Based on the feedback information, the preset scheduling rule is optimized, and based on the optimized preset scheduling rule, the scheduling strategy is optimized.

3. The heterogeneous task scheduling method for intelligent computing centers according to claim 2 is characterized in that: The optimizing the preset scheduling rule based on the feedback information, and optimizing the scheduling strategy based on the optimized preset scheduling rule, includes: Based on the feedback information, evaluating the execution effects of the heterogeneous tasks to obtain an evaluation result; Based on the evaluation result, the preset scheduling rule is optimized to obtain an optimized preset scheduling rule; Obtain updated computing resource information of each computing node; According to the optimized preset scheduling rules, based on the updated computing resource information, the task information and the computing resource requirements, the scheduling strategy is optimized to obtain an optimized scheduling strategy; Based on the optimized scheduling strategy, the heterogeneous tasks are scheduled.

4. The heterogeneous task scheduling method for intelligent computing centers according to claim 2 is characterized in that: The optimizing the preset scheduling rule includes at least one of the following: Optimizing the priority allocation algorithm; Optimizing the heuristic algorithm; Modify the preset load threshold of each computing power node.

5. The heterogeneous task scheduling method for intelligent computing centers according to claim 1 is characterized in that: The generating of a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to the preset scheduling rules includes: Screening the computing power nodes to obtain multiple candidate computing power nodes; Matching the computing power resource information of the multiple candidate computing power nodes with the task information and computing power resource requirements of the heterogeneous tasks to obtain a matching result; Based on the preset scheduling rules and the matching results, a scheduling strategy for each of the heterogeneous tasks is generated.

6. The heterogeneous task scheduling method for intelligent computing centers according to claim 1 is characterized in that: The heuristic algorithm is used to allocate computing resources to heterogeneous tasks, including: Construct an objective function with the goal of minimizing the total task execution time and / or maximizing the total resource utilization; Based on a heuristic algorithm, the objective function is solved to obtain a scheduling strategy for each heterogeneous task.

7. The heterogeneous task scheduling method for intelligent computing centers according to any one of claims 2 to 6, characterized in that: The computing power resource information includes the computing power resource type, the amount of computing power resources used and the amount of available computing power resources; the task information includes the task type and task indicators; the computing power resource demand includes the computing power resource demand and the computing power resource type demand.

8. A heterogeneous task scheduling system for intelligent computing centers, characterized by: include: Resource perception module, used to obtain computing resource information of each computing node in the intelligent computing center; Resource requirement assessment module, used to determine the task information and computing resource requirements of each heterogeneous task to be processed; A scheduling module is configured to generate a scheduling strategy based on the computing resource information, task information, and computing resource requirements according to preset scheduling rules, and schedule the heterogeneous tasks to corresponding target computing nodes based on the scheduling strategy; The preset scheduling rules include at least one of the following: Based on the priority allocation algorithm, computing resources are allocated to heterogeneous tasks; Based on heuristic algorithms, computing resources are allocated to heterogeneous tasks; Computing resources are allocated to heterogeneous tasks based on the preset load threshold of each computing node.

9. The heterogeneous task scheduling system for intelligent computing centers according to claim 8 is characterized in that: The system further comprises: A feedback module, configured to obtain feedback information after the execution of each heterogeneous task; The optimization module is used to optimize the preset scheduling rules based on the feedback information, and optimize the scheduling strategy based on the optimized preset scheduling rules.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, it implements the heterogeneous task scheduling method for the intelligent computing center as described in any one of claims 1 to 7.

Images