Resource scheduling method and device, electronic equipment and storage medium

Abstract

The invention provides a resource scheduling method and device, electronic equipment and a storage medium, relates to the technical field of data processing, and can solve the problems of low network resource utilization rate and low network performance. The method comprises the steps that service information is obtained according to a scheduling request sent by a terminal, a target scheduling algorithm matched with the service information is determined based on a scheduling algorithm model and a scheduling algorithm library, and the scheduling algorithm model is obtained based on real service information, simulation service information and the scheduling algorithm library training; the scheduling algorithm library comprises scheduling algorithms corresponding to a plurality of service scenes, the target scheduling algorithm is used for indicating a scheduling mode of network resource information matched with the service information, and the network resource information comprises cloud nodes, edge nodes and transmission paths; and scheduling the corresponding target processing node and the target transmission path for the service information. The network resource utilization rate and the network performance can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a resource scheduling method, device, electronic device and storage medium. Background Art

[0002] With the rapid development of network scale and the widespread popularity of network applications, the demand for large-scale computing in the network is increasing, and the traditional centralized computing architecture can no longer meet the current real-time and diversified task processing requirements.

[0003] At present, the idea of ​​cloud computing, network computing, edge computing and terminal devices working together has been proposed, but with the increase in the heterogeneity of computing nodes, the system complexity has increased significantly, resulting in low network resource utilization and network performance. Therefore, how to efficiently schedule resources at all levels to improve network resource utilization and network performance is an urgent problem to be solved. Summary of the invention

[0004] The present application provides a resource scheduling method, device, electronic device and storage medium, which can solve the problems of low network resource utilization and network performance.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a resource scheduling method, the method comprising: obtaining business information according to a scheduling request sent by a terminal, and determining a target scheduling algorithm that matches the business information based on a scheduling algorithm model and a scheduling algorithm library, wherein the scheduling algorithm model is obtained based on real business information, simulated business information, and training of a scheduling algorithm library, the scheduling algorithm library including scheduling algorithms corresponding to multiple business scenarios, the target scheduling algorithm being used to indicate a scheduling method for network resource information that matches the business information, the network resource information including cloud nodes, edge nodes, and transmission paths, and according to the target scheduling algorithm, scheduling a target processing node and a target transmission path corresponding to the business information.

[0007] The resource scheduling method provided in the embodiment of the present application can more accurately identify the business scenario corresponding to the current business information through a scheduling algorithm model trained based on real business information and simulated business information, ensure that the determined scheduling algorithm is closely integrated with the current business information, avoid waste of network resources, and improve the overall utilization efficiency of resources; at the same time, the precisely matched scheduling algorithm can more effectively process the traffic characteristics of different business types, ensure that resources such as cloud nodes, edge nodes, and transmission paths are efficiently utilized, reduce network congestion and transmission delays, and improve network performance. Therefore, the resource scheduling method provided in the embodiment of the present application can improve network resource utilization and network performance.

[0008] In combination with the above-mentioned first aspect, in a possible implementation method, the method also includes: obtaining first pre-training data, the first pre-training data including real business information, simulated business information, terminal parameter information and network resource information; dividing the first pre-training data into a training set, a verification set and a test set under different business scenarios; based on the training set and the verification set, determining the scheduling algorithms corresponding to different business scenarios; testing the scheduling performance of the scheduling algorithms corresponding to different business scenarios according to the test set, and storing the scheduling algorithms whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0009] In combination with the above-mentioned first aspect, in a possible implementation method, the above-mentioned scheduling algorithms corresponding to different business scenarios are determined based on the training set and the verification set, including: for each business scenario, multiple preset scheduling algorithms are used to schedule the training set to obtain the network resource energy consumption and data transmission delay corresponding to the business scenario; according to the network resource energy consumption and data transmission delay corresponding to the verification set and the business scenario, the corresponding network resource energy consumption difference and data transmission delay difference are determined after each preset scheduling algorithm schedules the training set; according to the network resource energy consumption difference and the data transmission delay difference, the scheduling algorithm corresponding to the business scenario is determined from multiple preset scheduling algorithms.

[0010] In combination with the above-mentioned first aspect, in a possible implementation method, the above-mentioned scheduling performance of the scheduling algorithm corresponding to different business scenarios is tested according to the test set, and the scheduling algorithm whose scheduling performance meets the preset performance is stored in the scheduling algorithm library, including: for each business scenario, when the network resource energy consumption difference is less than or equal to the preset resource energy consumption difference, and the data transmission delay difference is less than or equal to the preset transmission delay difference, determining that the scheduling performance meets the preset performance; and storing the scheduling algorithm whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0011] In combination with the above-mentioned first aspect, in a possible implementation method, the method also includes: obtaining second pre-training data, the second pre-training data including real business information, simulated business information, terminal parameter information, network resource information and scheduling algorithms corresponding to multiple business scenarios included in the scheduling algorithm library; dividing the second pre-training data into a training set, a verification set and a test set; using the training set to train the initial scheduling model to obtain the business scenario and scheduling algorithm corresponding to each business information; adjusting the parameters of the initial scheduling model according to the business scenario and scheduling algorithm corresponding to each business information in the verification set, and the business scenario and scheduling algorithm corresponding to each business information obtained by the initial scheduling model, to obtain an adjusted initial scheduling model; testing the accuracy of the adjusted initial scheduling model according to the test set, and determining the initial scheduling model whose accuracy meets the preset accuracy as the scheduling algorithm model.

[0012] In combination with the above-mentioned first aspect, in a possible implementation method, the accuracy of the initial scheduling model adjusted according to the test set test is determined as the scheduling algorithm model, including: when the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is greater than the first similarity, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is greater than the second similarity, determining that the accuracy meets the preset accuracy; and determining the initial scheduling model with the accuracy meeting the preset accuracy as the scheduling algorithm model.

[0013] In combination with the first aspect above, in a possible implementation method, the above-mentioned determination of the target scheduling algorithm that matches the business information based on the scheduling algorithm model and the scheduling algorithm library includes: inputting the business information into the scheduling algorithm model; identifying the target business scenario corresponding to the business information through the scheduling algorithm model, and determining the target scheduling algorithm corresponding to the target business scenario from the scheduling algorithm library; and outputting the target scheduling algorithm.

[0014] In the second aspect, the present application provides a resource scheduling device, which includes: an acquisition unit, a determination unit and a processing unit, wherein: the acquisition unit is used to acquire business information according to a scheduling request sent by a terminal; the determination unit is used to determine a target scheduling algorithm that matches the business information based on a scheduling algorithm model and a scheduling algorithm library, the scheduling algorithm model is based on real business information, simulated business information, and training of the scheduling algorithm library, the scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios, the target scheduling algorithm is used to indicate the scheduling method of network resource information that matches the business information, the network resource information includes cloud nodes, edge nodes and transmission paths; the processing unit is used to schedule the corresponding target processing nodes and target transmission paths for the business information according to the target scheduling algorithm.

[0015] In combination with the above-mentioned second aspect, in a possible implementation method, the acquisition unit is also used to obtain first pre-training data, the first pre-training data including real business information, simulated business information, terminal parameter information and network resource information; the processing unit is also used to divide the first pre-training data into training sets, verification sets and test sets under different business scenarios; the determination unit is also used to determine the scheduling algorithms corresponding to different business scenarios based on the training set and the verification set; the processing unit is also used to test the scheduling performance of the scheduling algorithms corresponding to different business scenarios according to the test set, and store the scheduling algorithms whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0016] In combination with the above-mentioned second aspect, in a possible implementation method, the determination unit is specifically used to: for each business scenario, use multiple preset scheduling algorithms to schedule the training set to obtain the network resource energy consumption and data transmission delay corresponding to the business scenario; determine the corresponding network resource energy consumption difference and data transmission delay difference after each preset scheduling algorithm schedules the training set based on the network resource energy consumption and data transmission delay difference; determine the scheduling algorithm corresponding to the business scenario from multiple preset scheduling algorithms based on the network resource energy consumption difference and data transmission delay difference.

[0017] In combination with the above-mentioned second aspect, in a possible implementation method, the processing unit is specifically used to: for each business scenario, when the network resource energy consumption difference is less than or equal to the preset resource energy consumption difference, and the data transmission delay difference is less than or equal to the preset transmission delay difference, determine that the scheduling performance meets the preset performance; and store the scheduling algorithm whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0018] In combination with the above-mentioned second aspect, in a possible implementation method, the acquisition unit is also used to: obtain second pre-training data, the second pre-training data includes real business information, simulated business information, terminal parameter information, network resource information, and scheduling algorithms corresponding to multiple business scenarios included in the scheduling algorithm library; the processing unit is also used to: divide the second pre-training data into a training set, a verification set, and a test set, use the training set to train the initial scheduling model to obtain the business scenario and scheduling algorithm corresponding to each business information, and adjust the parameters of the initial scheduling model according to the business scenario and scheduling algorithm corresponding to each business information in the verification set, and the business scenario and scheduling algorithm corresponding to each business information obtained by the initial scheduling model to obtain an adjusted initial scheduling model; the determination unit is also used to test the accuracy of the adjusted initial scheduling model according to the test set, and determine the initial scheduling model whose accuracy meets the preset accuracy as the scheduling algorithm model.

[0019] In combination with the above-mentioned second aspect, in a possible implementation method, the determination unit is specifically used to: determine that the accuracy rate meets the preset accuracy rate when the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is greater than the first similarity, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is greater than the second similarity; and determine the initial scheduling model whose accuracy meets the preset accuracy rate as the scheduling algorithm model.

[0020] In combination with the above-mentioned second aspect, in a possible implementation method, the determination unit is specifically used to: input business information into the scheduling algorithm model; identify the target business scenario corresponding to the business information through the scheduling algorithm model, and determine the target scheduling algorithm corresponding to the target business scenario from the scheduling algorithm library; and output the target scheduling algorithm.

[0021] In a third aspect, the present application provides an electronic device comprising: a processor and a communication interface; the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the resource scheduling method described in the first aspect and any possible implementation method of the first aspect.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program or instructions. When the computer program or instructions are executed on an electronic device, the electronic device executes the resource scheduling method described in the first aspect and any possible implementation of the first aspect.

[0023] In a fifth aspect, the present application provides a computer program product comprising a computer program or instructions. When the computer program product runs on an electronic device, the electronic device executes the resource scheduling method described in the first aspect and any possible implementation of the first aspect.

[0024] In a sixth aspect, the present application provides a chip, the chip comprising a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the resource scheduling method described in the first aspect and any possible implementation method of the first aspect.

[0025] Specifically, the chip provided in the present application also includes a memory for storing computer programs or instructions.

[0026] It should be noted that the above-mentioned computer program or instruction may be stored in whole or in part on a computer-readable storage medium, wherein the computer-readable storage medium may be packaged together with the processor of the device, or may be packaged separately from the processor of the device, which is not limited in this application.

[0027] In a seventh aspect, the present application provides a resource scheduling system, comprising: a resource scheduling device and a data server, wherein the resource scheduling device is used to execute the resource scheduling method described in the first aspect and any possible implementation manner of the first aspect.

[0028] The descriptions of the second to seventh aspects of the present application can refer to the detailed description of the first aspect; and the beneficial effects of the descriptions of the second to seventh aspects can refer to the beneficial effect analysis of the first aspect, which will not be repeated here.

[0029] In this application, the name of the resource scheduling device does not limit the device or functional module itself. In actual implementation, these devices or functional modules may appear with other names. As long as the functions of each device or functional module are similar to those of this application, they fall within the scope of the claims of this application and their equivalent technologies.

[0030] These and other aspects of the present application will become more apparent from the following description.

[0031] The above scheme brings at least the following beneficial effects: Based on the above technical scheme, the resource scheduling method provided by the present application can more accurately identify the business scenario corresponding to the current business information through the scheduling algorithm model trained based on real business information and simulated business information, ensure that the determined scheduling algorithm is closely integrated with the current business information, avoid waste of network resources, and improve the overall utilization efficiency of resources; at the same time, the precisely matched scheduling algorithm can more effectively process the traffic characteristics of different business types, ensure that resources such as cloud nodes, edge nodes and transmission paths are efficiently utilized, reduce network congestion and transmission delay, and improve network performance. Therefore, the resource scheduling method provided in the embodiment of the present application can improve network resource utilization and network performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A schematic diagram of the architecture of a resource scheduling system provided in an embodiment of the present application;

[0033] Figure 2 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application;

[0034] Figure 3 A flowchart of a resource scheduling method provided in an embodiment of the present application;

[0035] Figure 4 A flowchart of another resource scheduling method provided in an embodiment of the present application;

[0036] Figure 5 A training flow chart of a scheduling algorithm library provided in an embodiment of the present application;

[0037] Figure 6 A training flow chart of another scheduling algorithm library provided in an embodiment of the present application;

[0038] Figure 7 A training flow chart of a scheduling algorithm model provided in an embodiment of the present application;

[0039] Figure 8 A training flow chart of a genetic algorithm provided in an embodiment of the present application;

[0040] Fig. 9A schematic diagram of the structure of a resource scheduling device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0042] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0043] The terms "first" and "second" and the like in the specification and drawings of this application are used to distinguish different objects, or to distinguish different processing of the same object, rather than to describe a specific order of objects.

[0044] In addition, the terms "including" and "having" and any variations thereof mentioned in the description of the present application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0045] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0046] In the description of the present application, unless otherwise specified, “plurality” means two or more.

[0047] With the rapid development and widespread application of network technology, the demand for large-scale computing is increasing. The traditional centralized computing architecture can no longer meet the processing requirements of distributed, real-time and diversified tasks.

[0048] At present, the collaborative work of cloud computing, network computing, edge computing and terminal devices has gradually become the mainstream trend to meet this demand. The intelligent scheduling architecture of cloud-network-edge-end collaboration can effectively process, analyze and transmit massive data through the combination of cloud computing, network transmission and edge computing, thereby achieving real-time and intelligent business decisions, reducing the overall running time while performing high-performance computing and improving work efficiency.

[0049] Specifically, in the intelligent scheduling architecture of cloud-network-edge-end collaboration:

[0050] Cloud computing, as the core computing resource of the overall architecture, is responsible for high-density and high-complexity data tasks such as large-scale data storage, complex data analysis, and deep learning model training.

[0051] Edge computing makes full use of the advantage of being close to user terminals and is responsible for real-time processing, local analysis and temporary storage of data, alleviating the pressure of cloud computing.

[0052] Terminal devices are the forefront of the cloud-network-edge architecture. They interact directly with users and are responsible for task initiation and data collection.

[0053] The network layer, as the transmission medium connecting the cloud, edge, and end, is responsible for providing reliable bandwidth and low-latency transmission paths to ensure efficient data transmission and communication quality.

[0054] For example, the intelligent scheduling architecture that collaborates with the cloud, network, edge and end can be applied in the manufacturing industry to assist smart manufacturing and promote the automation, intelligence and flexibility of production lines; when applied in the medical and health field, it can improve the level of telemedicine, health monitoring and personalized diagnosis and treatment; and when applied in the construction of smart cities, it can realize the intelligence and refinement of urban management.

[0055] However, although cloud resources have powerful computing power and storage capacity, they are expensive and have low computing real-time performance, so it is necessary to avoid idle or overused resources; edge computing is close to the data source, has higher real-time performance and lower latency, but has relatively limited computing performance. It is suitable for processing tasks that are sensitive to latency and requires efficient coordination of resources between different edge nodes. Network resources are the key link between cloud resources, edge computing resources and terminal devices. Reasonable allocation of network resources (such as bandwidth and routing strategies) can improve the efficiency of data transmission, reduce latency and packet loss, and ensure the stability of real-time applications.

[0056] Therefore, how to achieve reasonable allocation and efficient scheduling of cloud resources, edge computing resources, and network resources while meeting task requirements to improve network performance and network resource utilization has become a key issue that needs to be solved urgently.

[0057] In the related technologies, the traditional strategies for cloud-network-edge resource scheduling usually design parameters based on specific business scenarios to achieve effective resource scheduling in specific business scenarios. However, they are not applicable to other different business scenarios and cannot achieve universal scheduling for all scenarios.

[0058] In some embodiments, factors that affect the cloud network edge resource scheduling effect include objective business and resources, and scheduling algorithms:

[0059] 1. Objective business and resources

[0060] 1) Different businesses have different requirements for computing resources, storage space, transmission latency, etc. For example, machine learning model training requires higher computing resources, while real-time video processing or autonomous driving requires lower latency. Existing scheduling strategies usually optimize the business of a certain scenario and cannot cope with the different needs of businesses in multiple scenarios.

[0061] 2) Different services have different priorities and different requirements for service quality (such as delay, bandwidth, and reliability). Existing scheduling strategies usually optimize a certain service and cannot handle services of different priorities, resulting in excessive resource occupation by low-priority services.

[0062] 3) The cloud, network, and edge layers have different types and computing power resources, such as different processor architectures (central processing unit (CPU), graphics processing unit (GPU), field programmable gate array (FPGA)), storage media (hard disk drive (HDD), solid state drive (SSD)), network bandwidth, etc., which makes the performance of each layer vary greatly. Existing scheduling strategies usually optimize a certain type of business and cannot effectively schedule different types of heterogeneous resources.

[0063] 4) There may be compatibility issues between the cloud, network layer, edge layer, and terminal devices. For example, the operating system or data format of the terminal device is different from that of the cloud, and additional conversion processes are required for data transmission and processing. The existing scheduling strategy cannot be adapted for compatibility, which may lead to low scheduling efficiency or system instability.

[0064] 2. Scheduling Algorithm

[0065] Traditional scheduling algorithms are mostly studied based on simulation platforms, but the simulation environment cannot reflect the actual situation of the network:

[0066] 1) Simulation data cannot fully reproduce the complexity of real situations, such as network topology, traffic patterns, and system load.

[0067] 2) Simulation data cannot show the information that already exists in real resource scheduling, such as geographical environment information, physical equipment information, and network topology information.

[0068] 3) Simulation data cannot reflect the actual computing cost of cloud-edge resources, resulting in poor matching between the resource scheduling strategy obtained based on simulation data and the actual network environment.

[0069] 4) Traditional scheduling algorithms only consider a single performance indicator (delay, energy consumption cost) during the scheduling process, and do not consider the comprehensive impact of multiple performance indicators.

[0070] For example, the polling algorithm only considers latency, so the latency indicator can be optimized by improving resource allocation. However, improving resource allocation costs too much and has a low cost-effectiveness. Therefore, collaborative optimization of multiple performance indicators can achieve optimal scheduling.

[0071] In addition, the optimization direction of services in different scenarios is also different. For example, scenarios with high real-time requirements, large-scale traffic scheduling, or terminal device empowerment all require different focuses for consideration and optimization.

[0072] Therefore, due to the diversity of services and resources, as well as the constraints of scheduling algorithms, the current scheduling strategies at the edge of the cloud network result in high actual scheduling complexity and difficulty, resulting in low network resource utilization and network performance.

[0073] In view of this, an embodiment of the present application provides a resource scheduling method that can be applied to the scenario of cloud-network-edge collaborative scheduling. Among them, the method can more accurately identify the business scenario corresponding to the current business information through a scheduling algorithm model trained based on real business information and simulated business information, ensure that the determined scheduling algorithm is closely integrated with the current business information, avoid waste of network resources, and improve the overall utilization efficiency of resources; at the same time, the precisely matched scheduling algorithm can more effectively handle the traffic characteristics of different business types, ensure that resources such as cloud nodes, edge nodes, and transmission paths are efficiently utilized, reduce network congestion and transmission delays, and improve network performance. Therefore, the resource scheduling method provided in the embodiment of the present application can improve network resource utilization and network performance.

[0074] The implementation of the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings.

[0075] Figure 1 This is a schematic diagram of the architecture of a resource scheduling system provided in an embodiment of the present application. Figure 1 As shown, the resource scheduling system 100 includes: a resource scheduling device 101 and a data server 102 .

[0076] The resource scheduling device 101 and the data server 102 may be one or more. For ease of understanding, Figure 1 Only one resource scheduling device and one data server are shown.

[0077] The resource scheduling device 101 and the data server 102 are connected via a communication link. The communication link may be a wired communication link or a wireless communication link, which is not limited in the present application.

[0078] In some embodiments, the resource scheduling device 101 can first obtain business information based on the scheduling request sent by the terminal; then determine the target scheduling algorithm that matches the business information based on the scheduling algorithm model and the scheduling algorithm library; finally, according to the target scheduling algorithm, schedule the corresponding target processing node and target transmission path for the business information.

[0079] Among them, the scheduling algorithm model is trained based on real business information, simulated business information, and a scheduling algorithm library. The scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios. The target scheduling algorithm is used to indicate the scheduling method of network resource information that matches the business information. The network resource information includes cloud nodes, edge nodes, and transmission paths.

[0080] In some embodiments, the data server 102 is used to store all scheduling algorithms in the scheduling algorithm library and to send a target scheduling algorithm matching the service information to the resource scheduling device 101. Accordingly, the resource scheduling device 101 receives the target scheduling algorithm sent by the data server 102.

[0081] When implemented by hardware, each module in the resource scheduling device 101 can be integrated into Figure 2 The hardware structure of the electronic device shown is implemented. Specifically, Figure 2 As shown, the basic hardware structure of electronic equipment is introduced.

[0082] Figure 2 The hardware structure diagram of an electronic device provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the electronic device includes at least one processor 201, a communication line 202, and at least one communication interface 204, and may also include a memory 203. The processor 201, the memory 203 and the communication interface 204 may be connected via the communication line 202.

[0083] The processor 201 may be a CPU, an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application, such as one or more digital signal processors (DSP), or one or more FPGAs.

[0084] The communication link 202 may include a pathway for transmitting information between the above-mentioned components.

[0085] The communication interface 204 is used to communicate with other devices or communication networks, and can use any transceiver-like device, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0086] The memory 203 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to include or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0087] In one possible design, the memory 203 can exist independently of the processor 201, that is, the memory 203 can be a memory outside the processor 201. In this case, the memory 203 can be connected to the processor 201 through the communication line 202, and is used to store execution instructions or application code, and the processor 201 controls the execution to implement the resource scheduling method provided in the following embodiment of the present application. In another possible design, the memory 203 can also be integrated with the processor 201, that is, the memory 203 can be the internal memory of the processor 201, for example, the memory 203 is a high-speed cache, which can be used to temporarily store some data and instruction information.

[0088] As a possible implementation, the processor 201 may include one or more CPUs, for example Figure 2 As another possible implementation, the electronic device may include multiple processors, such as Figure 2 As another possible implementation, the electronic device may further include an output device 205 and an input device 206.

[0089] It should be pointed out that the various embodiments of the present application can refer to each other, for example, the same or similar steps, method embodiments, system embodiments and device embodiments can refer to each other without limitation.

[0090] Figure 3 A flowchart of a resource scheduling method provided in an embodiment of the present application, the subject executing the method may be as follows Figure 1 The resource scheduling device 101 shown may also be various units / modules in the resource scheduling device 101, such as an integrated circuit or a chip, which is not specifically limited in the embodiment of the present application.

[0091] For example, Figure 3 As shown, the resource scheduling method provided in the embodiment of the present application includes the following steps:

[0092] S301. Obtain service information according to a scheduling request sent by a terminal.

[0093] Among them, the scheduling request is the terminal's request for cloud resources, edge computing resources and network resources; business information refers to the basic information of the business currently undertaken by the terminal.

[0094] In some embodiments, the service information includes service data and service requests. The service data includes service identification, terminal identification, service start time, service end time, resource request information, and service status information; the service request includes service identification, computing resources required for the service, network resources required for the service, and service quality requirements (such as delay, bandwidth).

[0095] S302: Determine a target scheduling algorithm that matches the service information based on the scheduling algorithm model and the scheduling algorithm library.

[0096] In some embodiments, the scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios.

[0097] Exemplarily, the scheduling algorithm library may include game theory algorithms, heuristic algorithms, polling algorithms, compromise algorithms, etc., wherein the heuristic algorithms may be further divided into: greedy algorithms, simulated annealing algorithms, and genetic algorithms.

[0098] The polling algorithm means that after the terminal initiates a service, the computing nodes are allocated according to the order in which the services are initiated.

[0099] The compromise algorithm means that after the terminal initiates a service, the network resource energy consumption required for the service is calculated according to the cost function, and the scheduling strategy with the minimum network resource energy consumption is selected.

[0100] Among them, the target scheduling algorithm is used to indicate the scheduling method of network resource information matching the business information, and the network resource information includes cloud nodes, edge nodes and transmission paths.

[0101] It should be noted that cloud nodes are cloud resources, edge nodes are edge computing resources, and transmission paths are network resources.

[0102] In some embodiments, after the business information is acquired, the business information may be preprocessed to obtain preprocessed business information so as to eliminate invalid data and abnormal data.

[0103] Exemplarily, the acquired business information may be cleansed to identify missing or invalid data points in the business information, interpolate the missing data points, and delete the invalid data points.

[0104] Optionally, after preprocessing the business information, the preprocessed business information can be input into the scheduling algorithm model, and then the scheduling algorithm model is used to identify the target business scenario corresponding to the business information, and the target scheduling algorithm corresponding to the target business scenario is determined from the scheduling algorithm library, and finally the target scheduling algorithm is output.

[0105] In some embodiments, since the scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios, after the business information is input into the scheduling algorithm model, the target scheduling algorithm corresponding to the business scenario can be screened from the scheduling algorithm library based on the business scenario corresponding to the business information.

[0106] In some embodiments, the scheduling algorithm model is trained based on real business information, simulated business information, and a scheduling algorithm library.

[0107] S303: Schedule corresponding target processing nodes and target transmission paths for the service information according to a target scheduling algorithm.

[0108] In some embodiments, network resource information may be scheduled according to a target scheduling algorithm determined by a scheduling algorithm model and service information to obtain a target processing node for processing the service information and a target transmission path for transmitting the service information.

[0109] Among them, network resource information includes network topology, network maximum bandwidth, network delay and network status.

[0110] Exemplarily, based on the target scheduling algorithm, the computing resources required for the business, the network resources required for the business, and the service quality requirements, the processing nodes (such as cloud node 1, edge node 1, edge node 2) corresponding to the computing resources required for the business can be determined from the network topology structure, and the network resources required for the business can be determined from the network bandwidth, and the delay and status that meet the service quality requirements of the business can be determined based on the network delay and network status.

[0111] Optionally, after the target processing node and the target transmission path corresponding to the service information are obtained through scheduling according to the target scheduling algorithm, the calculation result of the target processing node on the service information may also be output.

[0112] Exemplarily, it is taken that the target processing nodes include cloud node 1, edge node 1 and edge node 2. After cloud node 1, edge node 1 and edge node 2 complete the calculation of service information, the calculation results (network resource energy consumption, network transmission delay) can be output.

[0113] In the resource scheduling method provided in the embodiment of the present application, through the scheduling algorithm model trained based on real business information and simulated business information, the business scenario corresponding to the current business information can be more accurately identified, ensuring that the determined scheduling algorithm is closely integrated with the current business information, avoiding the waste of network resources, and improving the overall utilization efficiency of resources; at the same time, the precisely matched scheduling algorithm can more effectively handle the traffic characteristics of different business types, ensuring that resources such as cloud nodes, edge nodes, and transmission paths are efficiently utilized, reducing network congestion and transmission delays, and improving network performance. Therefore, the resource scheduling method provided in the embodiment of the present application can improve network resource utilization and network performance.

[0114] The following is an illustrative description of the complete process of the resource scheduling method provided in the embodiment of the present application.

[0115] For example, Figure 4 As shown, the complete process of the resource scheduling method may include the following steps:

[0116] S401. The terminal device initiates a service request.

[0117] S402: The resource scheduling device collects business information and performs pre-processing.

[0118] S403: The resource scheduling device determines the service performance requirement according to the service information.

[0119] S404: The resource scheduling device collects network resource information.

[0120] S405 , the resource scheduling device inputs the pre-processed service information, service performance requirements, and network resource information into the algorithm scheduling model.

[0121] S406: The algorithm scheduling model matches and outputs a target scheduling algorithm from a scheduling algorithm library.

[0122] S407: The resource scheduling device formulates a scheduling strategy according to the target scheduling algorithm.

[0123] S408. The resource scheduling device schedules the cloud network edge resources according to the scheduling strategy and outputs the calculation results.

[0124] As described in S302 above, in some embodiments, the scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios.

[0125] The following introduces the training process of the scheduling algorithm library in the embodiment of the present application.

[0126] For example, Figure 5 As shown, the training process of the scheduling algorithm library may include the following steps:

[0127] S501: Obtain first pre-training data.

[0128] The first pre-training data includes real business information, simulated business information, terminal parameter information and network resource information.

[0129] In some embodiments, real business information refers to business information obtained by capturing data of existing scenarios in a real network; simulated business information refers to building more business scenarios based on existing scenarios in a real network using a simulator to collect business information.

[0130] In some embodiments, the terminal parameter information includes device identification, device type, device location, device resource configuration, device implementation resources, and device status. The network resource information may refer to the description in S303 above.

[0131] Optionally, the first pre-training data may also include scheduling data information and workload data. The scheduling data information includes the scheduling algorithm and the priority rules adopted by the scheduling algorithm; the workload data includes the maximum number of CPUs actually running, the maximum memory actually occupied, the percentage of CPU, memory, and hard disk requested by the business, the percentage of network bandwidth requested by the business, the transmission delay of the business, and the energy consumption of network resources requested by the business after being scheduled by the scheduling algorithm.

[0132] S502: Divide the first pre-training data into a training set, a validation set, and a test set for different business scenarios.

[0133] Optionally, before dividing the data, the first pre-training data may be pre-processed to obtain pre-processed first pre-training data.

[0134] Exemplarily, different types of data can be formatted first to obtain first pre-trained data with a unified format, and then the first pre-trained data after formatting can be detected for missing values ​​and outliers, and the missing values ​​and outliers can be restored (such as deletion or interpolation), and then the data from different terminal sources can be normalized, and business information from the same source can be classified into the same batch.

[0135] Furthermore, the preprocessed first pre-training data may be subjected to scene annotation to identify the business scenario corresponding to each piece of business information, and divided into a training set, a validation set, and a test set according to a preset ratio.

[0136] For example, the first pre-training data after preprocessing includes 1000 sets of data. According to the ratio of training set 2: validation set 1: test set 1, 500 sets of data can be divided into training sets, 250 sets of data can be divided into validation sets, and 250 sets of data can be divided into test sets.

[0137] In some embodiments, data features that are strongly related to business scenarios can be obtained based on the common features and individual features of business information in different business scenarios.

[0138] Among them, common characteristics may include basic user information (age, gender), package usage (call charges, traffic), and network service quality (network speed, stability); individual characteristics may include specific bandwidth selection and specific delay selection.

[0139] Exemplarily, principal component analysis, multidimensional scaling analysis, convolution, filtering, etc. can be used to analyze the correlation between different features and the characteristics between features and target values ​​to obtain data features that are strongly related to the business scenario.

[0140] Principal component analysis is a statistical method for dimensionality reduction, which reduces the dimension and complexity of data by extracting the main components (i.e. common features) from the data.

[0141] Multidimensional scaling is a statistical method used to visualize high-dimensional data, which reveals the common characteristics between data by maintaining the relative distances between data points.

[0142] Convolution is a technology used in image processing and deep learning. It extracts features (common features and individual features) from data through convolution operations and is used for data classification, etc.

[0143] The filtering method is a feature selection method that filters out features that are irrelevant or redundant to the business scenario by setting certain conditions or thresholds, thereby retaining common or individual features that are strongly related to the business scenario.

[0144] S503: For each business scenario, multiple preset scheduling algorithms are used to schedule the training set to obtain network resource energy consumption and data transmission delay corresponding to the business scenario.

[0145] Among them, the multiple preset scheduling algorithms can be multiple scheduling algorithms set by default, and the scheduling efficiency of the preset scheduling algorithm is lower than the scheduling efficiency of the target scheduling algorithm.

[0146] In some embodiments, a training set can be used to train each preset scheduling algorithm for different business scenarios to obtain a trained scheduling algorithm, as well as network resource energy consumption and data transmission delay corresponding to the scheduling algorithm after each training.

[0147] S504. Determine the network resource energy consumption difference and data transmission delay difference corresponding to each preset scheduling algorithm after scheduling the training set according to the network resource energy consumption and data transmission delay corresponding to the verification set and the business scenario.

[0148] In some embodiments, the network resource energy consumption difference between the two can be determined based on the network resource energy consumption in the verification set and the network resource energy consumption corresponding to the scheduling algorithm after training; the data transmission delay difference between the two can also be determined based on the transmission delay in the verification set and the data transmission delay corresponding to the scheduling algorithm after training.

[0149] For example, taking the video processing scenario as an example, if the network resource energy consumption in the verification set is 10 and the data transmission delay is 2; the network resource energy consumption corresponding to the scheduling algorithm after training is 8 and the data transmission delay is 1, then the network resource energy consumption difference can be calculated to be 2 and the data transmission delay difference is 1.

[0150] S505. Determine a scheduling algorithm corresponding to the business scenario from a plurality of preset scheduling algorithms according to the network resource energy consumption difference and the data transmission delay difference.

[0151] In some embodiments, after obtaining multiple trained scheduling algorithms under the same business scenario and their corresponding network resource energy consumption differences and data transmission delay differences, the scheduling algorithm with the smallest network resource energy consumption difference and data transmission delay difference can be determined as the scheduling algorithm corresponding to the business scenario.

[0152] For example, take the three scheduling algorithms (algorithm A, algorithm B, and algorithm C) under the same business scenario as an example. If the network resource energy consumption difference corresponding to algorithm A is 2 and the data transmission delay difference is 1, if the network resource energy consumption difference corresponding to algorithm B is 0.8 and the data transmission delay difference is 0.5, if the network resource energy consumption difference corresponding to algorithm C is 1.5 and the data transmission delay difference is 1, then algorithm B can be determined as the scheduling algorithm corresponding to the business scenario.

[0153] S506. For each service scenario, when the network resource energy consumption difference is less than or equal to the preset resource energy consumption difference, and the data transmission delay difference is less than or equal to the preset transmission delay difference, determine that the scheduling performance meets the preset performance.

[0154] The preset resource energy consumption difference and the preset transmission delay difference can both be manually set values, which can be flexibly adjusted according to actual scenarios. For example, the preset resource energy consumption difference can be 1, and the preset transmission delay difference can be 0.8.

[0155] For example, the preset resource energy consumption difference is 1 and the preset transmission delay difference is 0.8. When the network resource energy consumption difference corresponding to algorithm B is 0.8 and the data transmission delay difference is 0.5, it indicates that the scheduling performance of algorithm B meets the preset performance.

[0156] S507: storing the scheduling algorithm whose scheduling performance meets the preset performance into the scheduling algorithm library.

[0157] In some embodiments, the scheduling algorithms that meet preset performance in various business scenarios can be summarized to construct a scheduling algorithm library that can be filtered based on the business scenarios.

[0158] In this way, multiple preset scheduling algorithms are trained through a large amount of training data to obtain the corresponding scheduling algorithm for each business scenario, and the corresponding scheduling algorithms for all business scenarios are constructed into a scheduling algorithm library that can be screened based on business scenarios, ensuring the accuracy of subsequent determination of the scheduling algorithm corresponding to the actual business information.

[0159] The following uses the genetic algorithm included in the scheduling algorithm library as an example to describe the definition of the objective function and constraints for maximizing the service success rate and minimizing the service delay:

[0160] 1) The overall maximum business success rate can be described by formula (1):

[0161]

[0162] Among them, x i，j The decision variable indicating whether the i-th task is successfully assigned to cloud node j, if assigned, it is 1, otherwise it is 0; N is the total number of tasks.

[0163] 2) The minimum service delay can be described by formula (2):

[0164]

[0165] Where S represents the set of delay-sensitive services, S i represents the data volume of the ith service, B li is the bandwidth of the selected transmission path li, Ci is the computing resource size required for the i-th business, C pi The computing speed of the cloud resources where the business is located.

[0166] 3) The objective function can be described by formula (III):

[0167] f(x)=max(αf1(x)-βf2(x)) Formula (III)

[0168] Among them, α is the weight coefficient of the service success rate priority (0≤α≤1); β=α-1 is the weight coefficient of the delay priority, and the objective function f(x) is the weighted sum of minimizing the service delay and maximizing the service success rate.

[0169] In some embodiments, the values ​​of α and β can be adjusted in real time according to business needs. In delay-sensitive scenarios, the system will increase the weight α to prioritize reducing business delays; in stability-first scenarios, the system will increase the weight β to ensure business success rate.

[0170] 4) Constraints may include the following ①, ②, ③, ④:

[0171] ①、 This means that each business can only be assigned to one cloud node;

[0172] ②、 Indicates that the amount of resources allocated to the cloud node cannot exceed the limit;

[0173] ③、 Indicates that the link allocated bandwidth cannot exceed the current link bandwidth limit;

[0174] ④、 Indicates that the service transmission delay cannot exceed the service's own delay requirement.

[0175] In this way, by optimizing the scheduling algorithm according to different business scenarios, it can be ensured that adaptive scheduling can be performed in the subsequent face of various business scenarios, thereby improving the availability of network resources after scheduling.

[0176] The following is an exemplary description of the complete training process of the scheduling algorithm library provided in the embodiment of the present application.

[0177] For example, Figure 6 As shown, the complete training process of the scheduling algorithm library may include the following steps:

[0178] S601. Collect service information and network resource information from the existing network and the simulation platform, pre-process the service information, and divide it into a training set, a verification set, and a test set.

[0179] S602: Analyze and extract data features that are strongly relevant to each business scenario.

[0180] S603: Label each piece of business information with a scenario, and perform training for each business scenario.

[0181] S604: For the training set under the same service scenario, different initial scheduling algorithms are used for scheduling, and the network resource energy consumption and data transmission delay after scheduling by each initial scheduling algorithm are calculated based on the verification set.

[0182] S605: Determine a target scheduling algorithm that matches the service scenario based on the network resource energy consumption and data transmission delay after scheduling by each initial scheduling algorithm.

[0183] S606: Test whether the scheduling performance of the target scheduling algorithm meets the preset performance based on the test set. If yes, execute S608; if no, execute S607.

[0184] S607: Adjust the parameters of the target scheduling algorithm and execute S604.

[0185] S608: Continue to train the target scheduling algorithm using the training set.

[0186] S609: Determine whether all training sets have been trained. If yes, execute S610; if no, execute S603.

[0187] S610: Summarize the target scheduling algorithms that match each business scenario and build a scheduling algorithm library that is screened based on the business scenario.

[0188] As described in S302 above, in some embodiments, the scheduling algorithm model is obtained by training based on real business information, simulated business information, and a scheduling algorithm library.

[0189] The following introduces the training process of the scheduling algorithm model in the embodiment of the present application.

[0190] For example, Figure 7 As shown, the training process of the scheduling algorithm model may include the following steps:

[0191] S701: Obtain second pre-training data.

[0192] The second pre-training data includes real business information, simulated business information, terminal parameter information, network resource information, and scheduling algorithms corresponding to multiple business scenarios included in the scheduling algorithm library.

[0193] Optionally, the second pre-training data may further include scheduling data information and workload data.

[0194] It should be noted that the detailed description of the second pre-training data may refer to the related description of the first pre-training data in S401 above, which will not be repeated here.

[0195] S702: Divide the second pre-training data into a training set, a validation set, and a test set.

[0196] Optionally, before dividing the data, the second pre-training data may be pre-processed to obtain pre-processed second pre-training data.

[0197] It should be noted that the preprocessing operation may refer to the preprocessing operation on the first pre-training data in S402 above, which will not be described in detail here.

[0198] For example, the pre-processed second pre-training data includes 1000 sets of data. According to the ratio of training set 2: validation set 1: test set 1, 500 sets of data are divided into training sets, 250 sets of data are divided into validation sets, and 250 sets of data are divided into test sets.

[0199] S703: Use the training set to train the initial scheduling model to obtain the business scenario and scheduling algorithm corresponding to each business information.

[0200] In some embodiments, the initial scheduling model may be an existing publicly available basic model, for example, the initial scheduling model may be a Qwen2-1.5B-Instruct model.

[0201] Optionally, before training the Qwen2-1.5B-Instruct model, the Qwen2-1.5B-Instruct model may be initialized to set initial parameters of the model.

[0202] For example, a deep learning framework (such as the transformers library) can be used to load model weights and configuration files, create a model instance, and then set the initial parameters of the model (such as learning rate, batch size, number of training rounds, etc.) according to specific task requirements.

[0203] Furthermore, the business information in the training set is input into the initialized Qwen2-1.5B-Instruct model in turn to obtain the business scenario and scheduling algorithm corresponding to each business information.

[0204] S704. According to the business scenario and scheduling algorithm corresponding to each business information in the verification set and the business scenario and scheduling algorithm corresponding to each business information obtained by the initial scheduling model, the parameters of the initial scheduling model are adjusted to obtain an adjusted initial scheduling model.

[0205] Among them, the scheduling accuracy of the adjusted initial scheduling model is greater than the scheduling accuracy of the initial scheduling model before adjustment.

[0206] In an optional implementation method, a first adjustment parameter can be obtained based on the business scenario corresponding to each business information in the verification set and the business scenario corresponding to each business information obtained by the initial scheduling model; a second adjustment parameter can also be obtained based on the scheduling algorithm corresponding to each business information in the verification set and the scheduling algorithm corresponding to each business information obtained by the initial scheduling model; and then the parameters of the initial scheduling model are adjusted based on the first adjustment parameter and the second adjustment parameter to obtain an adjusted initial scheduling model.

[0207] In another optional implementation method, a first adjustment parameter can be obtained based on the business scenario corresponding to each business information in the verification set and the business scenario corresponding to each business information obtained by the initial scheduling model; a second adjustment parameter can also be obtained based on the scheduling algorithm corresponding to each business information in the verification set and the scheduling algorithm corresponding to each business information obtained by the initial scheduling model; the first adjustment parameter and the second adjustment parameter are weighted averaged to obtain the target adjustment parameter, and finally the parameters of the initial scheduling model are adjusted according to the target adjustment parameter to obtain the adjusted initial scheduling model.

[0208] S705 . Test the accuracy of the adjusted initial scheduling model according to the test set, and determine the initial scheduling model whose accuracy meets the preset accuracy as the scheduling algorithm model.

[0209] The preset accuracy rate may be a manually set value, which may be flexibly adjusted according to actual scenarios. For example, the preset accuracy rate may be 95%.

[0210] In an optional implementation method, when the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is greater than the first similarity, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is greater than the second similarity, it is determined that the accuracy meets the preset accuracy, and the initial scheduling model whose accuracy meets the preset accuracy is determined as the scheduling algorithm model.

[0211] In some embodiments, the average value of the first similarity and the second similarity may be determined as the accuracy of the adjusted initial scheduling model.

[0212] The first similarity and the second similarity can both be manually set values, which can be flexibly adjusted according to actual scenarios. For example, the first similarity can be 95%, and the second similarity can be 97%.

[0213] Exemplarily, taking the first similarity as 95%, the second similarity as 97%, and the preset accuracy as 95%. If the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is 96%, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is 98%, then the accuracy of the adjusted initial scheduling model can be calculated to be (96%+98%) / 2=97%, which means that the accuracy of the adjusted initial scheduling model meets the preset accuracy, that is, the adjusted initial scheduling model can be determined as the final algorithm scheduling model.

[0214] In this way, the initial scheduling model is trained through a large amount of training data, and the accuracy of the initial scheduling model is continuously adjusted based on the verification set and the test set to ensure that the accuracy of the final algorithm scheduling model can meet the accuracy of subsequent use and improve the accuracy of the scheduling algorithm.

[0215] The following takes the above initial scheduling algorithm as a genetic algorithm as an example to illustrate the algorithm training process.

[0216] For example, Figure 8 As shown, the training process of the genetic algorithm may include the following steps:

[0217] S801: Obtain third pre-training data.

[0218] The third pre-training data may be the same as the first pre-training data or the second pre-training data.

[0219] Optionally, the service information table, the device location information table, the cloud node table, and the transmission link table may also be loaded according to the third pre-training data.

[0220] S802: Initialize genetic algorithm parameter configuration.

[0221] Exemplarily, the population size of the genetic parameter may be set to an initialization value of 50, and the mutation rate to 0.01.

[0222] Among them, the mutation rate refers to the probability of random change of data in each population.

[0223] S803: Initialize the population.

[0224] Each population represents a service allocation scheme, and there is an available link (network resource) between the cloud node and the terminal to which the service allocation scheme corresponds.

[0225] Exemplarily, the fitness of each population can be calculated by formula (IV) according to the maximum business success rate and resource utilization rate.

[0226]

[0227] Among them, S i For the number of successful transactions, For resource utilization.

[0228] S804: Determine whether the population fitness meets the termination condition. If not, execute S805; if yes, execute S808.

[0229] The termination condition may include fitness convergence or the number of iterations reaching a preset number of iterations.

[0230] In some embodiments, fitness convergence means that the fitness satisfies a preset fitness. The preset fitness may be a preset value, which may be flexibly adjusted according to different business scenarios.

[0231] S805: Determine whether the iteration reaches the maximum number of iterations. If yes, execute S807; if no, execute S806.

[0232] In some embodiments, the maximum number of iterations may be a specified number of times the algorithm runs. For example, the maximum number of iterations may be 100 times.

[0233] S806: Perform reselection, crossover, and mutation, and then execute S803.

[0234] Exemplarily, the selection methods may include the following Table 1, which only shows some selection methods (roulette selection, tournament selection, quadratic tournament selection (such as non-dominated sorting genetic algorithm (NSGA-II))), and actual use may also include more or fewer selection methods, which are not listed here one by one.

[0235] Table 1

[0236]

[0237] Exemplarily, the crossing methods may include those shown in Table 2 below. Table 2 only shows some of the crossing methods. In actual use, more or fewer crossing methods may be included, which are not listed one by one here.

[0238] Table 2

[0239]

[0240] S807: Output prompt information and scheduling strategy.

[0241] The prompt information may be that the performance indicator cannot be met.

[0242] S808. Output the scheduling strategy.

[0243] In this way, by continuously iteratively training the scheduling algorithm, the scheduling accuracy of the scheduling algorithm is improved, thereby indirectly improving the network performance and resource utilization after the scheduling algorithm schedules the network resources.

[0244] The embodiment of the present application can divide the resource scheduling device into functional modules or functional units according to the above method example. For example, each functional module or functional unit can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules or functional units. Among them, the division of modules or units in the embodiment of the present application is schematic, which is only a logical function division. There may be other division methods in actual implementation.

[0245] like Fig. 9 , which is a schematic diagram of the structure of a resource scheduling device 900 provided in an embodiment of the present application. The resource scheduling device 900 includes: an acquisition unit 901, a determination unit 902 and a processing unit 903.

[0246] Among them, the acquisition unit 901 is used to obtain business information according to the scheduling request sent by the terminal; the determination unit 902 is used to determine the target scheduling algorithm matching the business information based on the scheduling algorithm model and the scheduling algorithm library. The scheduling algorithm model is based on real business information, simulated business information, and the scheduling algorithm library is trained. The scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios. The target scheduling algorithm is used to indicate the scheduling method of network resource information matching the business information. The network resource information includes cloud nodes, edge nodes and transmission paths; the processing unit 903 is used to schedule the corresponding target processing nodes and target transmission paths for the business information according to the target scheduling algorithm.

[0247] In a possible implementation, the acquisition unit 901 is also used to obtain first pre-training data, the first pre-training data including real business information, simulated business information, terminal parameter information and network resource information; the processing unit 903 is also used to divide the first pre-training data into a training set, a verification set and a test set under different business scenarios; the determination unit 902 is also used to determine the scheduling algorithms corresponding to different business scenarios based on the training set and the verification set; the processing unit 903 is also used to test the scheduling performance of the scheduling algorithms corresponding to different business scenarios according to the test set, and store the scheduling algorithms whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0248] In one possible implementation, the determination unit 902 is specifically used to: for each business scenario, use multiple preset scheduling algorithms to schedule the training set to obtain the network resource energy consumption and data transmission delay corresponding to the business scenario; determine the corresponding network resource energy consumption difference and data transmission delay difference after each preset scheduling algorithm schedules the training set based on the network resource energy consumption and data transmission delay difference; determine the scheduling algorithm corresponding to the business scenario from multiple preset scheduling algorithms based on the network resource energy consumption difference and data transmission delay difference.

[0249] In one possible implementation, the processing unit 903 is specifically used to: for each business scenario, when the network resource energy consumption difference is less than or equal to the preset resource energy consumption difference, and the data transmission delay difference is less than or equal to the preset transmission delay difference, determine that the scheduling performance meets the preset performance; and store the scheduling algorithm whose scheduling performance meets the preset performance in the scheduling algorithm library.

[0250] In a possible implementation, the acquisition unit 901 is also used to: obtain second pre-training data, the second pre-training data includes real business information, simulated business information, terminal parameter information, network resource information and scheduling algorithms corresponding to multiple business scenarios included in the scheduling algorithm library; the processing unit 903 is also used to: divide the second pre-training data into a training set, a verification set and a test set, use the training set to train the initial scheduling model to obtain the business scenario and scheduling algorithm corresponding to each business information, adjust the parameters of the initial scheduling model according to the business scenario and scheduling algorithm corresponding to each business information in the verification set, and the business scenario and scheduling algorithm corresponding to each business information obtained by the initial scheduling model, to obtain an adjusted initial scheduling model; the determination unit 902 is also used to test the accuracy of the adjusted initial scheduling model according to the test set, and determine the initial scheduling model whose accuracy meets the preset accuracy as the scheduling algorithm model.

[0251] In one possible implementation, the determination unit 902 is specifically used to: determine that the accuracy meets the preset accuracy when the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is greater than the first similarity, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is greater than the second similarity; and determine the initial scheduling model whose accuracy meets the preset accuracy as the scheduling algorithm model.

[0252] In one possible implementation, determination unit 902 is specifically used to: input business information into a scheduling algorithm model; identify a target business scenario corresponding to the business information through the scheduling algorithm model, and determine a target scheduling algorithm corresponding to the target business scenario from a scheduling algorithm library; and output the target scheduling algorithm.

[0253] In the resource scheduling device provided in the embodiment of the present application, through the scheduling algorithm model trained based on real business information and simulated business information, it is possible to more accurately identify the business scenario corresponding to the current business information, ensure that the determined scheduling algorithm is closely integrated with the current business information, avoid the waste of network resources, and improve the overall utilization efficiency of resources; at the same time, the precisely matched scheduling algorithm can more effectively process the traffic characteristics of different business types, ensure that resources such as cloud nodes, edge nodes and transmission paths are efficiently utilized, reduce network congestion and transmission delay, and improve network performance. Therefore, the resource scheduling method provided in the embodiment of the present application can improve network resource utilization and network performance.

[0254] In a possible implementation, the resource scheduling device 900 may further include a storage unit 904 ( Fig. 9 The storage unit 904 stores a program or instruction. When the acquisition unit 901, the determination unit 902 and the processing unit 903 execute the program or instruction, the resource scheduling device 900 can execute the resource scheduling method described in the above method embodiment.

[0255] Through the description of the above implementation methods, technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0256] An embodiment of the present application provides a computer program product including instructions. When the computer program product is run on a computer, the computer is enabled to execute the resource scheduling method in the above method embodiment.

[0257] An embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a computer, the computer executes the resource scheduling method in the method flow shown in the above method embodiment.

[0258] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). In the embodiments of the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program, which may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0259] Since the resource scheduling device, computer-readable storage medium, and computer program product in the embodiments of the present application can be applied to the above method, the technical effects that can be obtained can also refer to the above method embodiments, and the embodiments of the present application will not be repeated here.

[0260] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0261] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of this disclosure can be achieved, and this application is not limited here.

[0262] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

[0263] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0264] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0265] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A resource scheduling method, characterized in that: The method comprises: Obtaining service information according to the scheduling request sent by the terminal; Based on a scheduling algorithm model and a scheduling algorithm library, determining a target scheduling algorithm matching the business information, wherein the scheduling algorithm model is obtained by training based on real business information, simulated business information, and the scheduling algorithm library, wherein the scheduling algorithm library includes scheduling algorithms corresponding to multiple business scenarios, respectively, and the target scheduling algorithm is used to indicate a scheduling method for network resource information matching the business information, wherein the network resource information includes cloud nodes, edge nodes, and transmission paths; According to the target scheduling algorithm, the corresponding target processing node and target transmission path are scheduled for the service information.

2. The method according to claim 1, characterized in that The method further comprises: Acquire first pre-training data, where the first pre-training data includes the real service information, the simulated service information, the terminal parameter information, and the network resource information; Dividing the first pre-training data into a training set, a validation set, and a test set for different business scenarios; Determining scheduling algorithms corresponding to different business scenarios based on the training set and the verification set; The scheduling performance of the scheduling algorithms corresponding to different business scenarios is tested according to the test set, and the scheduling algorithms whose scheduling performance meets the preset performance are stored in the scheduling algorithm library.

3. The method according to claim 2, characterized in that The determining, based on the training set and the verification set, a scheduling algorithm corresponding to different business scenarios includes: For each business scenario, multiple preset scheduling algorithms are used to schedule the training set to obtain network resource energy consumption and data transmission delay corresponding to the business scenario; According to the network resource energy consumption and data transmission delay corresponding to the verification set and the business scenario, determine the corresponding network resource energy consumption difference and data transmission delay difference after each preset scheduling algorithm schedules the training set; According to the network resource energy consumption difference and the data transmission delay difference, a scheduling algorithm corresponding to the business scenario is determined from the multiple preset scheduling algorithms.

4. The method according to claim 3, characterized in that The testing the scheduling performance of the scheduling algorithms corresponding to different business scenarios according to the test set, and storing the scheduling algorithms whose scheduling performance meets the preset performance in the scheduling algorithm library, includes: For each service scenario, when the network resource energy consumption difference is less than or equal to the preset resource energy consumption difference, and the data transmission delay difference is less than or equal to the preset transmission delay difference, determining that the scheduling performance meets the preset performance; The scheduling algorithm whose scheduling performance meets the preset performance is stored in the scheduling algorithm library.

5. The method according to claim 1, characterized in that The method further comprises: Acquire second pre-training data, where the second pre-training data includes the real service information, the simulated service information, the terminal parameter information, the network resource information, and a scheduling algorithm corresponding to each of a plurality of service scenarios included in the scheduling algorithm library; Dividing the second pre-training data into a training set, a validation set, and a test set; The initial scheduling model is trained using the training set to obtain a business scenario and a scheduling algorithm corresponding to each business information; According to the business scenario and scheduling algorithm corresponding to each business information in the verification set, and the business scenario and scheduling algorithm corresponding to each business information obtained by the initial scheduling model, adjusting the parameters of the initial scheduling model to obtain an adjusted initial scheduling model; The accuracy of the adjusted initial scheduling model is tested according to the test set, and the initial scheduling model whose accuracy meets the preset accuracy is determined as the scheduling algorithm model.

6. The method according to claim 5, characterized in that The step of testing the accuracy of the adjusted initial scheduling model according to the test set, and determining the initial scheduling model whose accuracy meets a preset accuracy as the scheduling algorithm model, includes: When the similarity between the business scenario corresponding to each business information obtained by the adjusted initial scheduling model and the business scenario corresponding to each business information in the test set is greater than the first similarity, and the similarity between the scheduling algorithm corresponding to each business information obtained by the adjusted initial scheduling model and the scheduling algorithm corresponding to each business information in the test set is greater than the second similarity, it is determined that the accuracy rate meets the preset accuracy rate; The initial scheduling model whose accuracy meets the preset accuracy is determined as the scheduling algorithm model.

7. The method according to any one of claims 1 to 6, characterized in that: The determining, based on the scheduling algorithm model and the scheduling algorithm library, a target scheduling algorithm matching the service information comprises: Inputting the service information into the scheduling algorithm model; Identifying the target business scenario corresponding to the business information through the scheduling algorithm model, and determining the target scheduling algorithm corresponding to the target business scenario from the scheduling algorithm library; The target scheduling algorithm is output.

8. A resource scheduling device, characterized in that: The device comprises: An acquisition unit, used to acquire service information according to a scheduling request sent by a terminal; a determination unit, configured to determine a target scheduling algorithm matching the service information based on a scheduling algorithm model and a scheduling algorithm library, wherein the scheduling algorithm model is obtained by training based on real service information, simulated service information, and the scheduling algorithm library, wherein the scheduling algorithm library includes scheduling algorithms corresponding to multiple service scenarios, respectively, and the target scheduling algorithm is used to indicate a scheduling method of network resource information matching the service information, wherein the network resource information includes cloud nodes, edge nodes, and transmission paths; The processing unit is used to schedule the corresponding target processing node and target transmission path for the service information according to the target scheduling algorithm.

9. An electronic device, characterized in that: include: A processor and a communication interface; the communication interface is coupled to the processor, and the processor is used to run a computer program or instruction to implement the resource scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program or instruction. When the electronic device executes the computer program or instruction, the electronic device executes the resource scheduling method according to any one of claims 1 to 7.

11. A computer program product, characterized in that The computer program product includes a computer program or instructions, and when the computer program or instructions are executed on an electronic device, the electronic device executes the resource scheduling method according to any one of claims 1 to 7.

Images