An Algorithm-Network Task Scheduling Method and Related Device Based on Improved Three-Way Decisions

Abstract

The present application discloses an arithmetic network task scheduling method and related devices based on improved three-way decisions, which relates to the technical fields of computing power resources and arithmetic network task scheduling. The method includes obtaining arithmetic network task scheduling data; classifying arithmetic network metrics according to the arithmetic network task scheduling data to construct an arithmetic network metric system; based on the arithmetic network metric system, using the three-way decision theory to divide each platform and each task respectively to obtain a platform division result and a task division result; according to the platform division result and the task division result, matching corresponding tasks to each platform to obtain a preliminary arithmetic network task scheduling result, and optimizing the preliminary arithmetic network task scheduling result according to the full load rate and matching degree of each platform to obtain a final arithmetic network task scheduling result. The present application can improve the utilization rate of arithmetic network resources and the efficiency of task scheduling.

Description

Technical Field

[0001] The present application relates to the technical field of computing power resources and computing network task scheduling, and in particular to a computing network task scheduling method and related devices based on improved three-way decision making. Background Art

[0002] In recent years, the widespread application of digital technologies such as artificial intelligence, blockchain, and the Internet of Things has led to a surge in data in the whole society, rapidly increasing the demand for data storage, computing, transmission, and application. The rise of computing power has become a new productivity, and the computing network, as a new architecture, aims to optimize business paths through the coordinated scheduling of distributed computing nodes, and maximize the user experience, network resources, and computing resource utilization efficiency. However, as a major challenge, the scheduling of computing network tasks requires solving how to intelligently and reasonably allocate tasks to different computing and network resources to achieve the goals of improving service quality, shortening task completion time, and achieving load balancing. How to efficiently process data with complexity, large scale, and spatiotemporal dynamics has become the focus of today's attention. In the design of many computing network task scheduling algorithms, tasks and resources are generally divided based on two-branch decision-making. However, for data with information uncertainty, traditional two-branch decision-making is difficult to provide an effective solution strategy, task scheduling efficiency is low, and computing network resource utilization is not high.

[0003] For example, in life, when information is uncertain, people's choices often include acceptance, rejection, and delayed decision-making. This is the theory of three-way decision. When dealing with such uncertainty, the three-way decision theory can provide an effective solution. The three-way decision is a decision-making model based on human cognition. It means that in the actual decision-making process, people can make quick judgments immediately for things that they are fully confident to accept or reject; for those things that cannot be decided immediately, people tend to postpone their judgments on events, that is, delayed decisions. The three-way decision theory has been successfully applied in the fields of medicine, engineering, management, and information. With the continuous deepening of the theoretical research on computing power networks, the value of the three-way decision theory in computing network applications has gradually emerged. Drawing on the theory and results of the three-way decision, we can better study the scheduling of computing network tasks, improve the utilization of computing network resources, and improve the efficiency of task scheduling, meeting the needs of cross-cloud and cross-region computing network resource scheduling.

[0004] Based on this, how to provide a computing network task scheduling method based on improved three-way decision-making to improve the computing network resource utilization and task scheduling efficiency has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0005] The objective of this application is to provide a computing and network task scheduling method and related devices based on improved three-way decisions, which can improve the utilization rate of computing and network resources and the efficiency of task scheduling.

[0006] To achieve the above objective, this application provides the following solutions:

[0007] In the first aspect, this application provides a computing and network task scheduling method based on improved three-way decisions. The computing and network task scheduling method based on improved three-way decisions includes:

[0008] Obtain computing and network task scheduling data; the computing and network task scheduling data includes the computing power information and network information of each platform, as well as the computing power information and network information of each task.

[0009] According to the computing and network task scheduling data, classify computing and network metrics and construct a computing and network metric system; the computing and network metric system includes several computing and network metrics. The computing and network metrics include computing power management metrics, computing power performance metrics, network capability metrics, and network performance metrics. The computing power management metrics refer to the metrics used to evaluate the management level of computing power providers. The computing power performance metrics refer to the metrics used to evaluate the performance of the computing power resources used by services. The network capability metrics refer to the metrics used to evaluate the functions and response capabilities of the network. The network performance metrics refer to the metrics used to evaluate the quality and stability of network transmission.

[0010] Based on the computing and network metric system, use the three-way decision theory to divide each platform and each task respectively, and obtain a platform division result and a task division result; the platform division result includes a computing-power strong platform, a network-capability strong platform, and a balanced platform. The computing-power strong platform refers to a platform where the computing power performance is stronger than the network performance. The network-capability strong platform refers to a platform where the computing power performance is weaker than the network performance. The balanced platform refers to a platform where the computing power performance is equal to the network performance. The task division result includes a computing-power strong task, a network-capability strong task, and a balanced task. The computing-power strong task refers to a task where the demand for computing power resources is more than the demand for network resources. The network-capability strong task refers to a task where the demand for computing power resources is weaker than the demand for network resources. The balanced task refers to a task where the demand for computing power resources is equal to the demand for network resources.

[0011] According to the platform division result and the task division result, match corresponding tasks to each platform to obtain a preliminary computing and network task scheduling result, and optimize the preliminary computing and network task scheduling result according to the full-load rate and matching degree of each platform to obtain a final computing and network task scheduling result; the final computing and network task scheduling result is used to represent the best scheduling result corresponding to allocating each task to the most suitable platform for execution.

[0012] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the computing-network task scheduling method based on improved three-way decisions described in any one of the above.

[0013] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the computing-network task scheduling method based on improved three-way decisions described in any one of the above.

[0014] In a fourth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the computing-network task scheduling method based on improved three-way decisions described in any one of the above.

[0015] According to the specific embodiments provided by the present application, the following technical effects are disclosed by the present application:

[0016] The present application provides a computing-network task scheduling method based on improved three-way decisions and related devices. By dividing computing-network metrics into computing power management metrics, computing power performance metrics, network capability metrics, and network performance metrics, a computing-network metric system is constructed. Then, based on the computing-network metric system, using the three-way decision theory, the platforms are divided into three categories: computing-power strong platforms, network-power strong platforms, and balanced platforms, and at the same time, the tasks are divided into three categories: computing-power strong tasks, network-power strong tasks, and balanced tasks, which helps to allocate different types of tasks to different types of platforms, realizes the targeted matching between different types of platforms and tasks, and further can complete the computing-network task scheduling more quickly and conveniently, improving the efficiency of computing-network task scheduling. Moreover, after the present application classifies and preliminarily schedules the platforms and tasks based on the three-way decision theory, it starts from two dimensions of the full-load rate and matching degree of the platforms, and optimizes the preliminary computing-network task scheduling results according to the full-load rate and matching degree of each platform after the preliminary scheduling, so as to ensure that each task is assigned to the most suitable platform for execution, obtain the best computing-network task scheduling results, and further effectively improve the utilization rate of computing-network resources. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is an application environment diagram of a computing-network task scheduling method based on improved three-way decisions provided by an embodiment of the present application.

[0019] Figure 2 It is a schematic flowchart of a computing and network task scheduling method based on improved three-way decisions provided by an embodiment of the present application.

[0020] Figure 3 It is a schematic diagram of the content of computing and network task scheduling based on improved three-way decisions provided by an embodiment of the present application.

[0021] Figure 4 It is a schematic diagram of the computing and network index system provided by an embodiment of the present application.

[0022] Figure 5 It is a schematic flowchart of the type division of the platform and tasks provided by an embodiment of the present application.

[0023] Figure 6 It is a schematic flowchart of the task scheduling process provided by an embodiment of the present application.

[0024] Figure 7 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0025] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0026] For the field of computing power resource scheduling and computing and network task scheduling, the accurate implementation of computing and network task scheduling can not only better adapt to the changes in the dynamic environment and tasks, but also achieve more refined task allocation and have better resource optimization potential. However, the existing computing and network task scheduling methods are mainly based on binary decision classification, ignoring the complexity of task scheduling brought by information uncertainty, resulting in resource waste and insufficient decision-making flexibility. In addition, the existing task scheduling methods still have problems such as the lack of construction of a standardized index system, fuzzy task classification, and lack of flexibility in scheduling, mainly focusing on the following three aspects:

[0027] (1) In terms of the construction of computing and network indexes: The existing computing and network task scheduling methods ignore the importance of constructing unified indexes, and the measurement schemes are single and one-sided, that is, they do not fully cover the current mainstream computing power and network performance indexes, resulting in the lack of key data and unreasonable index construction.

[0028] (2) In terms of computing-network task classification: Existing computing-network task scheduling methods mostly classify computing-network tasks based on traditional binary decision methods and complete task scheduling on this basis. However, in practical applications, the classification of computing-network tasks is usually not all binary classifications where it's either one or the other, and there is a phenomenon of an intermediate fuzzy task set, resulting in limitations of traditional binary decision methods in terms of the fine-grainedness of classification selection and resource utilization.

[0029] (3) In terms of computing-network task scheduling decision-making: Existing computing-network task scheduling methods still have problems such as low matching accuracy and poor flexibility, leading to unbalanced resource utilization and lack of elasticity in scheduling, ultimately reducing the accuracy of computing-network task scheduling.

[0030] Based on this, the technical objective of this application is to propose a computing-network task scheduling method based on improved three-way decisions, which is applied to the technical field of computing power resource scheduling and computing-network task scheduling, aiming to solve the above technical problems existing in the existing task scheduling methods and improve the utilization rate of computing-network resources and task scheduling efficiency.

[0031] To make the objectives, features, and advantages of this application more obvious and understandable, the following further details this application in combination with the accompanying drawings and specific embodiments.

[0032] A computing-network task scheduling method based on improved three-way decisions provided by an embodiment of this application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, placed on the cloud or other servers. The terminal 102 can send the computing and network task scheduling data to the server 104. After receiving the computing and network task scheduling data, for the computing and network task scheduling data, the server 104 classifies the computing and network metrics according to the computing and network task scheduling data, and constructs a computing and network metric system; based on the computing and network metric system, the three-way decision theory is used to divide each platform and each task respectively to obtain a platform division result and a task division result; according to the platform division result and the task division result, corresponding tasks are matched for each platform to obtain a preliminary computing and network task scheduling result, and the preliminary computing and network task scheduling result is optimized according to the full load rate and matching degree of each platform to obtain a final computing and network task scheduling result. The server 104 can feedback the obtained final computing and network task scheduling result to the terminal 102. In addition, in some embodiments, the computing and network task scheduling method based on the improved three-way decision can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly perform calculation and scheduling processing on the computing and network task scheduling data and output the computing and network task scheduling result, or the server 104 can obtain the computing and network task scheduling data from the data storage system and perform calculation and scheduling processing on the computing and network task scheduling data and output the computing and network task scheduling result.

[0033] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0034] In an exemplary embodiment, as Figure 2 shown, a computing and network task scheduling method based on the improved three-way decision is provided. This method is executed by a computer device, and can be specifically executed by a computer device such as a terminal or a server alone, or jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to Figure 1 the server 104 in

[0035] Step S1, obtain the computing and network task scheduling data.

[0036] In this embodiment, the computing and network task scheduling data includes the computing power information and network information of each platform, etc., and the computing power information and network information of each task, etc.

[0037] Step S2: Classify computing and network metrics according to the computing and network task scheduling data, and construct a computing and network metric system.

[0038] In this embodiment, the computing and network metric system includes a number of computing and network metrics. The computing and network metrics include computing power management metrics, computing power performance metrics, network capability metrics, and network performance metrics. The computing power management metrics refer to the metrics used to evaluate the management level of computing power providers. The computing power performance metrics refer to the metrics used to evaluate the performance of the computing power resources used by the service. The network capability metrics refer to the metrics used to evaluate the functions and response capabilities of the network. The network performance metrics refer to the metrics used to evaluate the quality and stability of network transmission.

[0039] In this embodiment, the computing power management metrics include service status, deployment location, cloud service provider type, and cloud service provider region, etc. The computing power performance metrics include turbo frequency, cache, video memory, CPU (Central Processing Unit) main frequency, number of CPU cores, number of CPU threads, average CPU load, CPU usage rate, memory capacity, memory usage rate, storage capacity, storage usage rate, storage IOPS (Input / Output Operations Per Second), GPU (Graphics Processing Unit) usage rate, FPGA (Field Programmable Gate Array) usage rate, ASCI (American Standard Code for Information Interchange) usage rate, and disk IO (Input / Output) usage rate, etc. The network capability metrics include average confirmation time, average recovery time, and reliability, etc. The network performance metrics include bandwidth, packet loss rate, end-to-end delay, handover delay, handover success rate, jitter, and bit rate, etc.

[0040] Step S3: Based on the computing and network metric system, use the three-way decision theory to divide each platform and each task respectively, and obtain the platform division result and the task division result.

[0041] In this embodiment, the platform division result includes three categories: computing-power-strong platforms, network-power-strong platforms, and balanced platforms. Among them, the computing-power-strong platforms refer to the platforms where the computing power performance is stronger than the network performance. The network-power-strong platforms refer to the platforms where the computing power performance is weaker than the network performance. The balanced platforms refer to the platforms where the computing power performance is equal to the network performance. The task division result includes three categories: computing-power-strong tasks, network-power-strong tasks, and balanced tasks. Among them, the computing-power-strong tasks refer to the tasks where the computing power resource requirements are more than the network resource requirements. The network-power-strong tasks refer to the tasks where the computing power resource requirements are weaker than the network resource requirements. The balanced tasks refer to the tasks where the computing power resource requirements are equal to the network resource requirements.

[0042] In this embodiment, step S3 specifically includes the following steps:

[0043] Step S31: According to the computing and network index system, using the method of Euclidean distance calculation, quantify each computing and network index respectively, and calculate the computing performance quantization values and network performance quantization values of each platform, as well as the computing performance quantization values and network performance quantization values of each task.

[0044] Step S32: Determine the magnitude relationship between the computing performance and network performance corresponding to each platform according to the computing performance quantization values and network performance quantization values of each platform.

[0045] Step S33: Determine the magnitude relationship between the computing resource requirements and network resource requirements corresponding to each task according to the computing performance quantization values and network performance quantization values of each task.

[0046] Step S34: According to the magnitude relationship between the computing performance and network performance corresponding to each platform, and the magnitude relationship between the computing resource requirements and network resource requirements corresponding to each task, use the three-way decision theory to divide each platform and each task respectively, and obtain the platform division result and task division result.

[0047] Step S4: According to the platform division result and the task division result, match corresponding tasks to each platform to obtain a preliminary computing and network task scheduling result, and optimize the preliminary computing and network task scheduling result according to the full load rate and matching degree of each platform to obtain the final computing and network task scheduling result. Among them, the final computing and network task scheduling result is used to represent the best scheduling result corresponding to allocating each task to the most suitable platform for execution.

[0048] In this embodiment, step S4 specifically includes the following steps:

[0049] Step S41: According to the platform division result and the task division result, calculate the matching degrees between each computing-strong platform and each computing-strong task, between each balanced platform and each balanced task, and between each network-strong platform and each network-strong task respectively, to obtain the matching degree calculation result.

[0050] Step S42: According to the matching degree calculation result, allocate the task with the highest matching degree to each platform to obtain a preliminary computing and network task scheduling result.

[0051] Step S43: Calculate the full load rate of each platform according to the preliminary computing and network task scheduling result.

[0052] Step S44: Adopt a greedy strategy to determine whether the current full-load rate and matching degree of each platform are the optimal solutions. If they are the optimal solutions, output the preliminary computing-network task scheduling result as the final computing-network task scheduling result; otherwise, return to the operation in step S3, "Based on the computing-network metric system, use the three-way decision theory to divide each platform and each task respectively to obtain the platform division result and the task division result", reset the thresholds used in the three-way decision theory, and re-divide each platform and each task until the current full-load rate and matching degree of each platform are the optimal solutions.

[0053] To make the technical solution of this embodiment clearer, the following uses an example to illustrate the specific implementation process of this technical solution in detail.

[0054] As Figure 3 shown, the computing-network task scheduling method based on the improved three-way decision in this embodiment mainly consists of three parts: computing-network metric construction, platform and task division, and task scheduling, and specifically includes the following steps.

[0055] Step 1: Computing-network metric construction. Adopt a stepped and modular construction technology to classify the computing-network metrics and construct a standardized computing-network metric system, including computing power management metrics, computing power performance metrics, network capability metrics, and network performance metrics.

[0056] In this embodiment, the computing-network metric construction adopts a stepped and modular construction technology, starting from two perspectives of computing power information and network information, and on this basis, it is further divided into computing power management metrics, computing power performance metrics, network capability metrics, and network performance metrics, with the characteristics of flexible combination, dynamic adjustment, and hierarchical systemization. Among them, the constructed computing-network metric system is as Figure 4 shown. The specific computing-network metrics of different types are as follows:

[0057] Computing power management metrics: Used to evaluate the relevant information of computing power providers to ensure the authenticity of computing power resources, and can be further divided into service status, deployment location, cloud service provider type, cloud service provider region, etc.

[0058] Computing power performance metrics: Used to evaluate the computing power resources used by the service, and can be further divided into turbo frequency, cache, video memory, CPU information (main frequency, number of cores, number of threads, average load, utilization rate), memory information (capacity, utilization rate), storage information (capacity, utilization rate, IOPS), utilization rate (GPU, FPGA, ASCI, disk IO), etc.

[0059] Network capability metrics: Used to evaluate the functions and response capabilities of the network to ensure that the computing-network can provide stable services in various situations, and can be further divided into average confirmation time, average recovery time, reliability, etc.

[0060] Network performance metrics: Used to evaluate the quality and stability of network transmission to ensure that data can be transmitted accurately and in a timely manner, and can be subdivided into bandwidth, packet loss rate, end-to-end delay, handover delay, handover success rate, jitter, rate / bit rate, etc.

[0061] Step 2: Platform and task division. Use three-way decisions to divide the existing platforms and tasks. Among them, the platforms are divided into platforms with computing power performance stronger than network performance (strong computing platforms), platforms with computing power performance weaker than network performance (strong network platforms), and platforms with computing power performance equal to network performance (balanced platforms); the tasks are divided into tasks with more computing power resource requirements than network resource requirements (strong computing tasks), tasks with weaker computing power resource requirements than network resource requirements (strong network tasks), and tasks with computing power resource requirements equal to network resource requirements (balanced tasks).

[0062] It should be noted that in practical applications, the balanced platform not only includes platforms with computing power performance equal to network performance, but also can include platforms with computing power performance close to network performance. Similarly, the balanced task not only includes tasks with computing power resource requirements equal to network resource requirements, but also can include tasks with computing power resource requirements close to network resource requirements. Here, "close" means that the difference between computing power performance and network performance, and between computing power resource requirements and network resource requirements is less than a certain preset threshold range, and the preset threshold can be set according to the actual situation.

[0063] In this embodiment, when dividing the platforms and tasks, on the one hand, it is considered that the computing power and network performance of different platforms have different characteristics; on the other hand, it is considered that different tasks have differences in computing power and network performance, and there are fuzzy characteristics in classification. Therefore, the platforms and tasks are further subdivided on the basis of the constructed computing-network index system. First, according to the Euclidean distance, the computing power and network performance of the platforms and tasks are quantified respectively; secondly, the platforms and tasks are classified through three-way decisions and threshold design, that is, the platforms are subdivided into strong computing platforms, strong network platforms and balanced platforms, and the tasks are subdivided into strong computing tasks, strong network tasks and balanced tasks. Finally, the division of tasks and platforms is completed. The specific steps of the platform and task division are as Figure 5 shown and include the following steps.

[0064] (1) In index quantification, it is known that the computing power platform P = {p1, p2,..., p m},where the computing power and network performance of the platform are expressed as p i = {p comi , p neti}, the computing power performance (p comi ) consists of computing power management metrics (Mp comi ) and computing power performance metrics (Pp comi ), and the network performance of the platform (p neti)Composed of network capability - type metrics (Mp neti ) and network performance - type metrics (Pp neti ). The task set is represented as T = {t1, t2,..., t n}, where the computing power and network resource requirements of the tasks are t i = {t comi , t neti}, t comi represents the computing power resources requested by the task, t neti represents the network resources requested by the task. The computing power performance of the task (t comi ) is composed of computing power management - type metrics (Mt comi ) and computing power performance - type metrics (Pt comi ). The network performance of the task (t neti ) is composed of network capability - type metrics (Mt neti ) and network performance - type metrics (Pt neti ). Normalize all computing power and network metrics and use the Euclidean distance calculation method. The calculation formulas are as shown in (1) and (2).

[0065] In this embodiment, the quantization formulas for the platform's computing power and network metrics are as follows:

[0066]

[0067] Among them, represents the quantization value of the platform's computing power performance, represents the quantization value of the platform's network performance, ω p1 represents the weight of the platform's computing power management - type metrics, ω p2 represents the weight of the platform's computing power performance - type metrics, ω p3 represents the weight of the platform's network capability - type metrics, ω p4 represents the weight of the platform's network performance - type metrics, ω p1 + ω p2 = 1, ω p3 + ω p4 = 1, represents the quantization value of the j - th computing power management - type metric under the i - th platform, represents the quantization value of the j - th computing power performance - type metric under the i - th platform, represents the quantization value of the j - th network management - type metric under the i - th platform, represents the quantization value of the j - th network performance - type metric under the i - th platform.

[0068] Similarly, the quantization formulas for the computing power and network metrics of different tasks can be obtained as follows:

[0069]

[0070] Among them, represents the quantitative value of the computing power performance of the task, represents the quantitative value of the network performance of the task, ω t1 represents the weight of the computing power management type index of the task, ω t2 represents the weight of the computing power performance type index of the task, ω t3 represents the weight of the network capability type index of the task, ω t4 represents the weight of the network performance type index of the task, ω t1 +ω t2 = 1, ω t3 +ω t4 = 1, represents the quantitative value of the j-th computing power management type index under the i-th task, represents the quantitative value of the j-th computing power performance type index under the i-th task, represents the quantitative value of the j-th network management type index under the i-th task, represents the quantitative value of the j-th network performance type index under the i-th task.

[0071] (2) In classification, using the three-way decision idea, the platform and tasks are classified by setting thresholds. The expression for platform classification is P = POS(P) ∪ BND(P) ∪ NEG(P), which satisfies the following properties:

[0072] 1) The three types of platforms have no intersection:

[0073] 2) The three types of platforms cover all platforms: POS(P) ∪ BND(P) ∪ NEG(P) = P.

[0074] The representation of the three-way decision for each platform in this embodiment is as follows:

[0075]

[0076] Among them, α1 and β1 are the thresholds for platform type classification, and 0 ≤ α1 < 0.5 < β1 ≤ 1. P represents the platform set, and POS(P), BND(P), and NEG(P) respectively correspond to the aforementioned 3 cases, namely the platform with stronger computing power performance than network performance (strong computing power platform), the platform with equal computing power performance and network performance (balanced platform), and the platform with weaker computing power performance than network performance (strong network platform). The union of the three is the total number of all platforms. For each domain, allocate appropriate tasks for execution.

[0077] Similarly, the expression for task classification is T = POS(T) ∪ BND(T) ∪ NEG(T), which satisfies the following properties:

[0078] 1) The three types of tasks have no intersection:

[0079] 2) The three types of tasks cover all tasks: POS(T) ∪ BND(T) ∪ NEG(T) = T.

[0080] The representation of the three-way decisions for each task in this embodiment is as follows:

[0081]

[0082] Among them, α2 and β2 are the thresholds for task type division, which are the same as the thresholds for platform type division and will be continuously adjusted and updated later, and 0 ≤ α2 < 0.5 < β2 ≤ 1. T represents the task set, and POS(T), BND(T), and NEG(T) respectively correspond to the aforementioned 3 situations, that is, tasks with more computing power resource requirements than network resource requirements (strong computing tasks), tasks with equal computing power resource requirements and network resource requirements (balanced tasks), and tasks with weaker computing power resource requirements than network resource requirements (strong network tasks). For each domain, tasks are assigned to appropriate platforms for execution to achieve task scheduling.

[0083] Step 3: Task scheduling. Perform platform matching on tasks, and complete the optimal matching of tasks and platforms by setting the optimal platform full load rate and matching degree adjustment threshold, so as to achieve precise task scheduling.

[0084] In this embodiment, during task scheduling, classification of platforms and tasks is achieved according to the initially set thresholds, and different types of tasks are assigned to corresponding platforms. Considering the dynamic nature of tasks and the requirements of fine-grained scheduling strategies, the threshold setting is dynamically adjusted according to the optimal platform full load rate to optimize system resources and achieve precise task scheduling. The specific steps are as follows.

[0085] (1) Calculate the matching degrees of the classified platforms and tasks respectively, as shown in formula (5). Assign the task with the highest platform matching degree to the corresponding platform, and subtract the resources consumed by the task from the platform, waiting for the next matching.

[0086]

[0087] Among them, represents the matching degree of task t i and platform p i . In the matching degree calculation, since the classification of tasks and platforms was achieved in the previous step, it is not necessary to calculate the matching degrees of all tasks and all platforms, that is, calculate the similarity between strong computing tasks and strong computing platforms, the similarity between strong network tasks and strong network platforms, and the similarity between balanced tasks and balanced platforms, which reduces the computational complexity and improves the matching performance.

[0088] (2) After the initial matching of tasks is completed, calculate the full load rate of each type of platform, as shown in formula (6).

[0089]

[0090] Among them, respectively represent the overall full load rates of computing-power-strong platforms, balanced platforms, and network-power-strong platforms. g, k, and q are the numbers of computing-power-strong platforms, balanced platforms, and network-power-strong platforms respectively. represents the computing power resources already used in the platform. represents the network resources already used in the platform, and this remaining amount is determined by the tasks that have been allocated.

[0091] (3) According to the distribution of the full load rates and matching degrees of computing-power-strong platforms, balanced platforms, and network-power-strong platforms, combined with the greedy strategy, perform dynamic threshold adjustment. Among them, the calculation methods of the full load rate and matching degree of the platform based on the greedy strategy are iterated with the full load rate of different types of platforms as the greedy object, and the tasks are mapped to the appropriate platforms. After each task allocation is completed, the task list will be reduced. Each iteration will generate the current best calculation results of the full load rate and matching degree, and update the threshold. When all iterations are completed, the current global allocation method is an approximate optimal solution. Finally, the reasonable scheduling of tasks is realized, ensuring the full utilization of platform resources.

[0092] As Figure 6 shown, the execution steps of the task scheduling method in this embodiment are as follows:

[0093] 1) Matching degree calculation. According to the classified platforms and tasks, calculate the platform matching degree, and schedule the tasks to the platform with the highest matching degree.

[0094] 2) Full load rate calculation. When the task allocation is completed, calculate the full load rates of different platforms.

[0095] 3) Optimal solution determination. Based on the greedy strategy, determine whether the current full load rate and matching degree reach the optimum. If they reach the optimum, jump to step 5); otherwise, jump to step 4).

[0096] 4) Threshold update. Update the thresholds (i.e., α1, β1, α2, and β2) set based on the three-way decision idea, and jump to step 1).

[0097] 5) Output the computing-network task scheduling strategy and complete the computing-network task scheduling.

[0098] This embodiment adopts a stepped and modular computing-network metric system construction method. Based on the stepped and modular construction technology, the computing-network metrics are constructed from two perspectives of computing power information and network information. The metrics are subdivided into computing power management metrics, computing power performance metrics, network capacity metrics, and network performance metrics, comprehensively considering various aspects of computing-network metrics to ensure the comprehensiveness of the metrics. Moreover, in this embodiment, according to the Euclidean distance, by using the Euclidean distance to quantify the computing power and network performance of the platform and the task, combined with the idea of three-way decision-making and threshold design, the task and the platform capabilities are respectively divided into three categories. The platform is subdivided into a computing-power-strong platform, a network-capability-strong platform, and a balanced platform, and the task is subdivided into a computing-power-strong task, a network-capability-strong task, and a balanced task. Furthermore, in this embodiment, the matching degree calculation is also used to schedule the task to the platform with the highest matching degree. Through the greedy strategy, the optimal task matching degree and the platform full-load rate are solved; the threshold is dynamically updated to obtain the optimal task scheduling strategy and complete the task scheduling. By subdividing the computing power and network metrics, a multi-dimensional and comprehensive metric system is constructed; through the classification module, accurate classification of the platform and the task is realized; through the task scheduling module, the matching degree is optimized to complete the accurate and dynamic scheduling of the task, thus solving the deficiencies of the existing computing-network task scheduling methods in aspects such as unified metric construction, the ambiguity of binary decision-making, and scheduling dynamics, overcoming problems such as a single measurement scheme, coarse-grained task classification, and fixed task scheduling methods, and providing a more flexible, comprehensive, and accurate solution for computing-network task scheduling.

[0099] This embodiment provides a computing-network task scheduling method based on improved three-way decision-making. First, the stepped and modular construction technology is adopted to flexibly combine, dynamically adjust, and hierarchically systemize the construction of computing-network metrics from two perspectives of computing power information and network information. The metrics are divided into computing power management metrics, computing power performance metrics, network capacity metrics, and network performance metrics, realizing a more detailed computing-network metric system. Secondly, combined with the idea of three-way decision-making, the traditional binary decision-making method is improved, and the task and the platform are subdivided into three categories, namely, tasks with more computing power resource requirements than network resource requirements, tasks with weaker computing power resource requirements than network resource requirements, tasks with equal computing power resource requirements and network resource requirements, platforms with more computing power resource requirements than network resource requirements, platforms with weaker computing power resource requirements than network resource requirements, and platforms with equal computing power resource requirements and network resource requirements. Finally, according to the full-load rate of the platform and the task matching degree, the classification threshold is dynamically adjusted to enhance the flexibility of the scheduling strategy and realize a more fine-grained overall strategy selection. The comprehensive improvement of the above three aspects will effectively improve the adaptability and flexibility of computing-network task scheduling in complex and dynamic environments, optimize system resources, and enhance the stability of system performance.

[0100] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data for computing network task scheduling. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a computing network task scheduling method based on improved three-way decisions.

[0101] Those skilled in the art can understand that Figure 7 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0102] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, it implements the steps of the above-mentioned computing network task scheduling method based on improved three-way decisions.

[0103] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by the processor, it implements the steps of the above-mentioned computing network task scheduling method based on improved three-way decisions.

[0104] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, it implements the steps of the above-mentioned computing network task scheduling method based on improved three-way decisions.

[0105] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0106] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0107] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0108] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for scheduling computing network tasks based on improved three-branch decision making, characterized in that: The computing network task scheduling method based on improved three-branch decision-making includes: Obtaining computing network task scheduling data; the computing network task scheduling data includes computing power information and network information of each platform and computing power information and network information of each task; According to the computing network task scheduling data, the computing network indicators are classified to construct a computing network indicator system; the computing network indicator system includes several computing network indicators, and the computing network indicators include computing power management indicators, computing power performance indicators, network capability indicators and network performance indicators. The computing power management indicators refer to indicators used to evaluate the management level of computing power providers, the computing power performance indicators refer to indicators used to evaluate the performance of computing power resources used by services, the network capability indicators refer to indicators used to evaluate the functions and responsiveness of the network, and the network performance indicators refer to indicators used to evaluate the quality and stability of network transmission; Based on the computing network indicator system, the three-branch decision theory is used to divide each platform and each task respectively, and the platform division results and task division results are obtained; the platform division results include a strong computing platform, a strong network platform and a balanced platform, the strong computing platform refers to a platform with computing power performance stronger than network performance, the strong network platform refers to a platform with computing power performance weaker than network performance, and the balanced platform refers to a platform with computing power performance equal to network performance; the task division results include strong computing tasks, strong network tasks and balanced tasks, the strong computing tasks refer to tasks with computing power resource requirements greater than network resource requirements, the strong network tasks refer to tasks with computing power resource requirements weaker than network resource requirements, and the balanced tasks refer to tasks with computing power resource requirements equal to network resource requirements; According to the platform division results and the task division results, the corresponding tasks are matched for each platform to obtain a preliminary network computing task scheduling result, and according to the full load rate and matching degree of each platform, the preliminary network computing task scheduling result is optimized to obtain the final network computing task scheduling result; the final network computing task scheduling result is used to represent the best scheduling result corresponding to allocating each task to the most suitable platform for execution.

2. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 1 is characterized in that: Based on the computing network indicator system, the three-branch decision theory is used to divide each platform and each task respectively, and the platform division results and task division results are obtained, which specifically include: According to the computing network indicator system, the Euclidean distance calculation method is used to quantify each of the computing network indicators, and the computing power performance quantification value and network performance quantification value of each platform, as well as the computing power performance quantification value and network performance quantification value of each task are calculated; According to the quantified values ​​of computing power performance and network performance of each platform, determine the size relationship between the computing power performance and network performance of each platform; According to the quantified values ​​of computing power performance and network performance of each task, determine the size relationship between the computing power resource requirements and network resource requirements corresponding to each task; According to the relationship between the computing power performance and network performance corresponding to each platform, as well as the relationship between the computing power resource requirements and network resource requirements corresponding to each task, the three-branch decision theory is used to divide each platform and each task respectively, and the platform division results and task division results are obtained.

3. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 2 is characterized in that: Use the following formula to calculate the quantified value of computing power performance and network performance of each platform: in, Indicates the quantified value of the computing power performance of the platform, p neti Represents the quantitative value of the platform's network performance, ω p1 Represents the weight of the platform computing power management indicators, ω p2 Indicates the weight of the platform computing power performance indicators, ω p3 Represents the weight of platform network capability indicators, ω p4 Indicates the weight of platform network performance indicators. It represents the quantitative value of the jth computing power management indicator under the i-th platform. It represents the quantitative value of the jth computing power performance indicator under the i-th platform. represents the quantitative value of the jth network management indicator under the i-th platform, It represents the quantitative value of the jth network performance indicator under the i-th platform; Use the following formula to calculate the computing power performance quantification value and network performance quantification value of each task: in, Indicates the quantified value of the computing power performance of the task, Represents the quantitative value of the network performance of the task, ω t1 Represents the weight of task computing power management indicators, ω t2 Represents the weight of the task computing performance indicator, ω t3 represents the weight of the task network capability indicator, ω t4 represents the weight of the task network performance indicator, It represents the quantitative value of the jth computing power management indicator under the i-th task. It represents the quantitative value of the jth computing power performance indicator under the i-th task. represents the quantitative value of the jth network management indicator under the i-th task, It represents the quantitative value of the j-th network performance indicator under the i-th task.

4. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 2 is characterized in that: The following formula is used to express the three-branch decision theory to divide each platform and obtain the platform division result: Among them, α1 and β1 are the thresholds for platform type classification, POS(P), BND(P) and NEG(P) represent computing-strong platform, balanced platform and network-strong platform respectively, and P represents the platform set. Indicates the quantified value of the computing power performance of the platform. Indicates the quantitative value of the platform's network performance; The following formula is used to express the division of each task using the three-branch decision theory to obtain the task division result: Among them, α2 and β2 are the thresholds for task type classification, POS(T), BND(T) and NEG(T) represent computing-intensive tasks, balancing tasks and network-intensive tasks respectively, and T represents the task set. Indicates the quantified value of the computing power performance of the task, A quantitative value representing the network performance of a task.

5. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 1 is characterized in that: According to the platform division results and the task division results, the corresponding tasks are matched for each platform to obtain a preliminary network computing task scheduling result, and according to the full load rate and matching degree of each platform, the preliminary network computing task scheduling result is optimized to obtain the final network computing task scheduling result, which specifically includes: According to the platform division results and the task division results, respectively calculate the matching degree between each computing power platform and each computing power task, the matching degree between each balancing platform and each balancing task, and the matching degree between each network strength platform and each network strength task to obtain a matching degree calculation result; According to the matching degree calculation result, the task with the highest matching degree is assigned to each platform to obtain a preliminary network computing task scheduling result; Calculate the full load rate of each platform according to the preliminary network computing task scheduling result; A greedy strategy is adopted to determine whether the current full load rate and matching degree of each platform are the optimal solution. If it is the optimal solution, the preliminary computing network task scheduling result is used as the final computing network task scheduling result. Otherwise, the process returns to the step of "based on the computing network indicator system, using the three-branch decision theory to divide each platform and each task respectively, and obtain the platform division result and the task division result", reset the threshold of the three-branch decision theory, and re-divide each platform and each task until the current full load rate and matching degree of each platform are the optimal solution.

6. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 5 is characterized in that: Use the following formula to calculate the full load factor of each platform: in, and They represent the overall full load rates of the strong computing platform, balanced platform, and strong network platform, respectively. g, k, and q are the numbers of strong computing platforms, balanced platforms, and strong network platforms, respectively. Indicates the computing resources used in the platform. Indicates the network resources used in the platform. Indicates the quantified value of the computing power performance of the platform. Indicates the quantitative value of the platform's network performance.

7. The method for scheduling network computing tasks based on improved three-branch decision making according to claim 1 is characterized in that: The computing power management indicators include service status, deployment location, cloud provider type and cloud provider region; The computing power performance indicators include turbo frequency, cache, video memory, CPU main frequency, number of CPU cores, number of CPU threads, CPU average load, CPU utilization, memory capacity, memory utilization, storage capacity, storage utilization, storage IOPS, GPU utilization, FPGA utilization, ASCI utilization and disk IO utilization; The network capability indicators include average confirmation time, average recovery time and reliability; The network performance indicators include bandwidth, packet loss rate, end-to-end delay, switching delay, switching success rate, jitter and bit rate.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the computing network task scheduling method based on improved three-way decision-making as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computing network task scheduling method based on improved three-way decision-making described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computing network task scheduling method based on improved three-way decision-making described in any one of claims 1 to 7 is implemented.

Images