Computing power resource scheduling method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of scheduling services, in particular to a computing power resource scheduling method and device, electronic equipment, a storage medium and a computer program product. The method comprises the steps of obtaining computing power node information from a storage device in response to a computing task request of a service request end; wherein the computing task request comprises a task parameter and a computing power resource performance requirement parameter; screening the computing power node information according to the computing task request, and determining available computing power node information; and sending the available computing power node information to the service request end. According to the method, more reasonable computing power resources can be allocated to the computing tasks, and reasonable scheduling of the computing power resources is realized.

Description

Technical Field

[0001] The present application relates to the technical field of scheduling services, and in particular, to a computing power resource scheduling method, apparatus, electronic device, storage medium, and computer program product. Background Art

[0002] A computing power network is a new computing power resource solution that can integrate edge computing nodes, cloud computing nodes, and various network resources including wide area networks. It is an infrastructure framework that can coordinate the computing and storage resources of cloud computing nodes and the network resources of wide area networks and provide flexible and schedulable on-demand services according to business characteristics. In existing computing power resource scheduling strategies, when scheduling computing power nodes, the computing power measurement of the computing power nodes is generally mainly considered. Summary of the Invention

[0003] Based on this, it is necessary to provide a resource scheduling method that can consider the performance requirements of computing power nodes during the computing power scheduling process to achieve reasonable scheduling of computing power resources for the above technical problems.

[0004] In a first aspect, the present application provides a computing power resource scheduling method, and the method includes: in response to a computing task request from a service request end, obtaining computing power node information from a storage device; wherein, the computing task request includes: task parameters and computing power resource performance requirement parameters;

[0005] Screening the computing power node information according to the computing task request to determine available computing power node information;

[0006] Sending the available computing power node information to the service request end.

[0007] In one embodiment, the method further includes: in response to a target trigger event, sending a node information update request to a computing power node; receiving the current node information fed back by the computing power node and saving it to the storage device for the storage device to update the computing power node information according to the current node information; and feeding back an update response information to the computing power node.

[0008] In one embodiment, the target trigger event includes: a trigger event corresponding to a node periodic update task, or a trigger event corresponding to the computing task request.

[0009] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0010] In one embodiment, the screening the computing power nodes in the computing power node information according to the computing task request to determine available computing power node information includes:

[0011] Take the task parameters and the computing power resource performance requirement parameters as input parameters and input them into the computing power resource analysis model to obtain available computing power node information;

[0012] The computing power resource analysis model is used to sort the computing power nodes to determine the computing power node sorting result; and output the available computing power node information based on the computing power node sorting result.

[0013] In one embodiment, the method further includes:

[0014] Use the computing power resource analysis model to generate a calculation task energy consumption prediction result based on the task parameters, and sort the computing power nodes in combination with the calculation task energy consumption prediction result.

[0015] In one embodiment, the available computing power node information includes: a first available computing power node and a second available computing power node;

[0016] The method further includes:

[0017] Determine the first available computing power node according to the computing power node sorting result;

[0018] When the reliability parameter of the first available computing power node is lower than the preset reliability standard, determine the second available computing power node according to the computing power node sorting result, so as to use the first available computing power node and the second available computing power node to execute the calculation task respectively.

[0019] In one embodiment, the node reliability requirement parameter includes at least one of a transmission capacity requirement parameter, a task priority parameter, and a node task completion rate parameter.

[0020] In one embodiment, the node energy consumption requirement parameter includes at least one of a node status requirement information, a node device type requirement information, a node energy consumption type requirement information, and a node location requirement information.

[0021] In one embodiment, the available computing power node information includes an energy consumption mode parameter and a processor frequency band parameter corresponding to the available computing power node.

[0022] In one embodiment, the method further includes:

[0023] In response to a software configuration request of the available computing power node for the calculation task, feedback the software data or the compiled software data corresponding to the calculation task to the available computing power node, so that the available computing power node uses the software data or the compiled software data to execute the calculation task.

[0024] In one embodiment, the method further includes:

[0025] Collect the node performance detection information corresponding to the available computing power nodes when executing the computing task;

[0026] When it is identified that the node performance detection information does not meet the preset node performance requirements, determine the standby computing power nodes according to the computing power node sorting result;

[0027] Send the standby computing power nodes to the service request end for the service request end to allocate the computing task to the standby computing power nodes, and the standby computing power nodes execute the computing task.

[0028] In one embodiment, the method further includes:

[0029] When it is identified that there are no available computing power nodes in the computing power node information, send a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

[0030] In one embodiment, the computing task request includes: device hardware requirement parameters; the method further includes:

[0031] Identify the computing power node information according to the device hardware requirement parameters, and when it is identified that there are computing power nodes that meet the device hardware requirement parameters, call a computing power resource analysis model to determine the available computing power node information; or

[0032] When it is identified that there are no computing power nodes that meet the device hardware requirement parameters, send a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

[0033] In one embodiment, the method further includes:

[0034] When the computing task corresponding to the computing task request is splittable, perform a splitting process on the computing task to obtain a computing subtask queue;

[0035] Based on the computing subtask queue, sequentially execute and configure the available computing power node information corresponding to each subtask.

[0036] In one embodiment, the method further includes:

[0037] Receive a node registration request of a computing power node; wherein, the node registration request includes node performance information corresponding to the computing power node; the node performance information includes a node reliability parameter and a node energy consumption parameter;

[0038] Save the computing power node to a storage device, and feedback a registration response message to the computing power node.

[0039] In a second aspect, the present application provides a computing power resource scheduling device, which includes:

[0040] A computing task request response module, configured to obtain computing power node information from a storage device in response to a computing task request from a service request end; wherein, the computing task request includes: task parameters and computing power resource performance requirement parameters;

[0041] A computing power node determination module, configured to screen the computing power node information according to the computing task request to determine available computing power node information;

[0042] A computing task allocation module, configured to send the available computing power node information to the service request end and screen the computing power node information according to the computing task request to determine available computing power node information.

[0043] In one embodiment, the device further includes:

[0044] A node information update module, configured to send a node information update request to a computing power node in response to a target trigger event to obtain the current node information fed back by the computing power node, and update the computing power node information of the storage device based on the current node information.

[0045] In one embodiment, the target trigger event includes: a trigger event corresponding to a node periodic update task, or a trigger event corresponding to the computing task request.

[0046] In one embodiment, the device further includes:

[0047] A software configuration request response module, configured to, in response to a software configuration request of an available computing power node for the computing task, feedback software data corresponding to the computing task or compiled software data to the available computing power node for the available computing power node to execute the computing task using the software data or the compiled software data.

[0048] In one embodiment, the device further includes:

[0049] A node performance monitoring module, configured to collect node performance detection information corresponding to the available computing power node when executing the computing task; and determine a standby computing power node according to the computing power node sorting result when identifying that the node performance detection information does not meet the preset node performance requirements, so as to execute the computing task using the standby computing power node.

[0050] In one embodiment, the device further includes:

[0051] An external computing power resource scheduling module, which is used to send a computing power resource scheduling request to an external computing power device when there is no available computing power node in the identified computing power node information, so as to obtain the allocated external computing power device as the available computing power node information.

[0052] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0053] In a third aspect, the present application also provides an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the computing power resource scheduling method described in any embodiment of the first aspect are implemented.

[0054] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the computing power resource scheduling method described in any embodiment of the first aspect are implemented.

[0055] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the computing power resource scheduling method described in any embodiment of the first aspect are implemented.

[0056] In a sixth aspect, the present application provides a computing power resource scheduling method, and the method includes:

[0057] Sending a computing task request to a computing power resource scheduling end; the computing task request includes task parameters and computing power resource performance requirement parameters;

[0058] Receiving the available computing power node information sent by the computing power resource scheduling end; the available computing power node information is determined by the computing power resource scheduling end according to the computing task request by screening the computing power node information;

[0059] Allocating the computing task to the computing power node corresponding to the available computing power node information, so as to obtain the computing task execution result fed back by the computing power node.

[0060] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0061] In one embodiment, the node reliability requirement parameters include at least one of the following: transmission capacity requirement parameters, task priority parameters, and node task completion rate parameters.

[0062] In one embodiment, the node energy consumption requirement parameters include at least one of: node status requirement information, node device type requirement information, node energy consumption type requirement information, and node location requirement information.

[0063] In one embodiment, the available computing power node information is determined by the computing power resource scheduling end screening the computing power node information according to the computing task request, and includes:

[0064] The computing power resource scheduling end uses the task parameters and the computing power resource performance requirement parameters as input parameters and inputs them into a computing power resource analysis model to obtain available computing power node information;

[0065] Among them, the computing power resource analysis model is used to sort the computing power nodes to determine the computing power node sorting result; and output available computing power node information based on the computing power node sorting result.

[0066] In one embodiment, the method further includes:

[0067] The computing power resource scheduling end uses the computing power resource analysis model to generate a computing task energy consumption prediction result based on the task parameters, and sorts the computing power nodes in combination with the computing task energy consumption prediction result.

[0068] In one embodiment, the available computing power node information includes: a first available computing power node and a second available computing power node;

[0069] The method further includes:

[0070] Allocate the computing task to the first available computing power node and the second available computing power node to respectively obtain the computing task execution results fed back by the first available computing power node and the second available computing power node when executing the computing task.

[0071] In one embodiment, the method further includes:

[0072] The computing power resource scheduling end determines a first available computing power node according to the computing power node sorting result; and when identifying that the reliability parameter of the first available computing power node is lower than a preset reliability standard, determines the second available computing power node according to the computing power node sorting result.

[0073] In one embodiment, the available computing power node information includes the energy consumption mode parameter and the processor frequency band parameter corresponding to the available computing power node.

[0074] In one embodiment, the computing task request includes: device hardware requirement parameters; the method further includes:

[0075] The computing power resource scheduling end identifies the computing power node information according to the device hardware requirement parameters, and when identifying that there is a computing power node that meets the device hardware requirement parameters, calls a computing power resource analysis model to determine the available computing power node information; or

[0076] When identifying that there is no computing power node that meets the device hardware requirement parameters, it sends a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

[0077] In one embodiment, the computing task includes a plurality of subtasks that can be split; the method further includes:

[0078] Receiving the available computing power node information corresponding to each subtask fed back by the computing power resource scheduling end.

[0079] In one embodiment, the method further includes:

[0080] Collecting the node performance detection information corresponding to the available computing power node when executing the computing task;

[0081] When identifying that the node performance detection information does not meet the preset node performance requirements, sending node exception information to the computing power resource scheduling end;

[0082] Obtaining the backup computing power node information sent by the computing power resource scheduling end, and allocating the computing task to the backup computing power node.

[0083] In a seventh aspect, the present application provides a computing power resource scheduling device, and the device includes:

[0084] A request processing module, configured to send a computing task request to a computing power resource scheduling end; the computing task request includes task parameters and computing power resource performance requirement parameters;

[0085] A node information receiving module, configured to receive the available computing power node information sent by the computing power resource scheduling end; the available computing power node information is determined by the computing power resource scheduling end according to the computing task request to screen the computing power node information;

[0086] A computing task allocation module, configured to allocate the computing task to the computing power node corresponding to the available computing power node information, so as to obtain the computing task execution result fed back by the computing power node.

[0087] In one embodiment, the available computing power node information includes: a first available computing power node, a second available computing power node; the computing task allocation module further includes:

[0088] A task reallocation module, configured to allocate the computing task to the first available computing power node and the second available computing power node, so as to respectively obtain the computing task execution results fed back by the first available computing power node and the second available computing power node when executing the computing task.

[0089] In one embodiment, the device further includes:

[0090] A computing power node reset module, configured to collect node performance detection information corresponding to the available computing power node when executing the computing task; when identifying that the node performance detection information does not meet the preset node performance requirements, send node exception information to the computing power resource scheduling end; so as to obtain the standby computing power node information sent by the computing power resource scheduling end, and allocate the computing task to the standby computing power node.

[0091] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0092] In a eighth aspect, the present application further provides an electronic device. The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the computing power resource scheduling method in any one of the embodiments of the sixth aspect are implemented.

[0093] In a ninth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the steps of the computing power resource scheduling method in any one of the embodiments of the sixth aspect are implemented.

[0094] In a tenth aspect, the present application further provides a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the computing power resource scheduling method in any one of the embodiments of the sixth aspect are implemented.

[0095] For the above-mentioned computing power resource scheduling method, after receiving the computing task request from the service request end, it can first obtain the computing power node information; then screen the computing power nodes according to the computing power resource performance requirement parameters and specific task parameters required in the computing task request, and determine the available computing power nodes that can execute the specific computing task, so as to fully consider the actual requirements for the performance of the self-care resources during the computing power resource scheduling process, and then be able to allocate more reasonable computing power resources for the computing task requested by the computing task request, and achieve reasonable scheduling of the computing power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 It is a schematic flowchart of the computing power resource scheduling method in one embodiment;

[0097] Figure 2 Schematic diagram of the system architecture for implementing the computing power resource scheduling method in an embodiment;

[0098] Figure 3 Interaction timing schematic diagram of the computing power resource scheduling method in an embodiment;

[0099] Figure 4 Flow schematic diagram of the computing power node registration method in an embodiment;

[0100] Figure 5 Flow schematic diagram of the method for updating computing power node information in an embodiment;

[0101] Figure 6 Flow schematic diagram of the method for monitoring the performance of the computing power node executing the computing task in an embodiment;

[0102] Figure 7 Flow schematic diagram of the computing power resource scheduling method of the computing power server in an embodiment;

[0103] Figure 8 Flow schematic diagram of another computing power resource scheduling method in an embodiment;

[0104] Figure 9 Structural block diagram of the computing power resource scheduling device in an embodiment;

[0105] Figure 10 Structural block diagram of another computing power resource scheduling device in an embodiment;

[0106] Figure 11 Internal structure diagram of the electronic device in an embodiment. Detailed implementation manners

[0107] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0108] In the related art, the computing nodes in the computing power network have the characteristics of diverse hardware devices and diverse connection methods. Different computing nodes generally have different computing capabilities, and there are also certain differences in the hardware environments of the hardware devices of different computing nodes. When scheduling computing power resources, generally only the computing capabilities of the computing nodes are considered, and other performances are lacking in consideration.

[0109] The computing power resource scheduling method provided by the embodiments of the present application can be applied to the system architecture as shown in Figure 2 shown. Refer to Figure 2As shown in the figure, a system architecture for a computing power resource scheduling method is provided, including a server 201, a data storage system 202, a terminal device, and a computing power network 206. Among them, the server 201 can be implemented by an independent server or a server cluster composed of multiple servers; the terminal device can be an intelligent electronic device such as a laptop 203, a mobile terminal 204, a desktop computer 205, etc.; the computing power network can include multiple computing power nodes, and each computing power node is used to execute specific computing tasks. The data storage system 202 can be used to store computing power node information. The server 201 communicates with the terminal device, the computing power network 206, and the data storage system through the network. The task request initiated by the terminal device can be sent to the server 201 through the network; the server 201, as the scheduling server, can deploy a computing power resource scheduling system to allocate computing power resources for the corresponding computing tasks according to the requests of the terminal device; and the specific computing tasks are executed by the computing power network 206.

[0110] In one embodiment, a computing power resource scheduling method is provided, which can be executed by the server side and is used to realize the orchestration and scheduling of computing power resources.

[0111] In one embodiment, referring to Figure 4 As shown in the figure, the computing power resource scheduling method may include:

[0112] Step S41, receiving a node registration request of a computing power node; wherein, the node registration request includes node performance information corresponding to the computing power node; the node performance information includes a node reliability parameter and a node energy consumption parameter;

[0113] Step S42, saving the computing power node to a storage device, and feeding back registration response information to the computing power node.

[0114] Specifically, for each computing power node in the computing power network, a node registration request can be sent to the scheduling server in advance, and the node information corresponding to the computing power node is carried in the request. Among them, the node information can include node performance information. After receiving the registration request of the computing power node, the scheduling server can save the corresponding computing power node information to the storage device and feed back the corresponding registration response information to the computing power node to notify the computing power node that the registration is successful. In the storage device, a computing power node list can be established to save the identity identifiers of each computing power node, as well as the specific node information and performance parameters.

[0115] Or, when the registration of the computing power node fails, the scheduling server can feed back a registration failure response information to the computing power node; and can trigger the computing power node to re-collect node information and re-initiate a registration request.

[0116] In one embodiment, in a storage device, the node information of a computing power node may include: basic node device information, node energy consumption parameters, and node reliability parameters.

[0117] Exemplarily, the basic node device information may include: device type, processor information, storage device information; data such as the location information of the computing power node. Among them, the device type can be used to describe the hardware device type of the computing power node. For example, the device type is a base station, a mobile terminal, an edge computing device MEC, etc.; in addition, it can also be marked whether the computing power node is a dedicated computing power resource device.

[0118] The above-mentioned processor information can be used to describe parameters such as the processor type, the number of cores, and the main frequency of the hardware device of the computing power node; for example, the processor type can be any one or more of: CPU (Central Processing Unit, central processor), GPU (graphics processing unit, graphics processor), TPU (tensor processing unit, tensor processor), ARM (Advanced RISC Machine, advanced reduced instruction set machine) architecture, FPGA (Field Programmable Gate Array, field programmable gate array). In addition, the processor information can also include information such as the current idle computing power of the processor that affects the computing power of the computing power node. The storage device information can be used to describe information such as the storage device type and available space of the computing power node; for example: the level and space of the cache (high-speed buffer memory); and the storage device type, read bandwidth, and expected idle storage size of other storage devices, etc., which affect the computing power. The location information can be used to describe the current specific location coordinates of the computing power node. In addition, for a computing power node of the mobile terminal type, it can also include the current movement state of the computing power node.

[0119] Exemplarily, the node reliability information may include any one or more of the following: transmission capacity parameter, task priority parameter, and node task completion rate parameter. Among them, the transmission capacity parameter can be used to describe the communication mode type and security parameter of the computing power node. Among them, the communication mode type may include wired and / or wireless connection modes; and the number of forwarding nodes. The security parameter can be used to describe the authentication mode supported by the computing power node and the information on the type / algorithm complexity of the supported security algorithm. The task priority parameter can be used to represent the priority level of the computing task on the computing power node; for example, in a multi-task sharing device, the task priority of the computing task in the device; for example, the computing task is the highest priority task of the computing power node; or it can also be the priority determined according to the occupancy ratio of the processor and memory, etc. The node task completion rate parameter can be used to represent the completion ratio of the historical computing tasks executed by the computing power node and the probability prediction result of the completion rate of subsequent computing tasks. Among them, the completion ratio of the historical computing tasks can be determined by statistically analyzing task information such as the completion situation, task type, and task failure reason of the historical tasks of the computing power node. In addition, the probability prediction result of the completion rate of subsequent computing tasks can be estimated based on the task type and data volume of the current computing task, combined with the historical task completion ratio of the computing power node and the failure reason of the failed tasks.

[0120] Exemplarily, the node energy consumption parameter may include any one or more combinations of the following: node status information, node energy consumption information, and node environment information. Among them, the node status information can be used to describe the current working state of the computing power node device, such as on, sleep, energy-saving, or off, abnormal, faulty, etc. In addition, the node status information can also include the power information of the node device, such as the current remaining power (unit: Kwh or mA*V), total power, etc. of the mobile terminal device. The node environment information can be used to describe the ambient temperature of the geographical location where the computing power node device is located, and whether cooling auxiliary devices such as air conditioners and fans are equipped; for example, when the ambient temperature of the node device is low, relatively cool, and no auxiliary cooling device is required, the energy consumption for refrigeration of the computing power node can be saved.

[0121] The node energy consumption type information can be used to describe the energy usage status and energy type of the computing power node device. Specifically, it can include the power limitation status, power saving requirement, and energy type of the computing power node device. Among them, the power limitation status of the node device can be used to indicate that the node device is a power-limited mobile terminal or mobile base station / TRP (multi-transmit and receive point), or it can also be a device without power limitation; the node device can be used to limit the type of computing tasks that can be processed and the corresponding task energy consumption according to the power limitation status. Among them, the power saving requirement can be used to describe the current power saving requirement of the node device; for example, it can be a power saving mode triggered when the remaining power of a node device of the mobile terminal type is insufficient; or, it can be a normal usage mode when the node device has sufficient current power or is connected to a power source. Among them, the energy type can be used to describe whether the energy type used by the node device is green energy; for example, whether it is solar energy; or, it can also describe the energy price used by the node device, whether it is high-price energy or low-price energy.

[0122] In one embodiment, as Figure 1 shown, the computing power resource scheduling method may include the following steps:

[0123] Step S11, in response to the computing task request from the service request side, obtain the computing power node information from the storage device; wherein, the computing task request includes: task parameters, computing power resource performance requirement parameters.

[0124] Exemplarily, for the scheduling server, it can receive computing task requests from various terminal devices. The computing task request can be request information generated by the service request side according to the current specific computing task. Among them, the computing task request can carry the task parameters corresponding to the computing task and the performance requirements for the computing power resources of the computing task, that is, the computing power resource performance requirement parameters. After the scheduling server receives the computing task request from the terminal device, it can trigger a node information extraction task and execute the node information extraction task to extract the computing power resource node information from the storage device. Among them, for the storage device, it can pre-store the node information corresponding to each computing power node in the computing power network; the node information can be used to describe the computing power, performance, energy consumption, etc. of each computing node.

[0125] Step S12, screen the computing power node information according to the computing task request to determine the available computing power node information.

[0126] Exemplarily, on the scheduling server side, after obtaining the computing node information of each computing node from the storage device, the computing nodes can be screened and sorted according to the reliability requirements, energy-saving requirements for the computing nodes in the computing task, and the specific computing power requirements of the computing task, so as to determine the available computing nodes for the current computing task request. Among them, the available computing node information may include the specific information of one or more computing nodes, and these computing nodes can be used to execute the computing task corresponding to the computing task request. For the scheduling server side, the computing nodes can be screened and sorted through a pre-trained computing power resource analysis model, and the available computing node information can be determined.

[0127] Step S13: Send the available computing node information to the service request side.

[0128] Exemplarily, on the scheduling server side, after determining the available computing node information, the available computing node information can be fed back to the service request side. As a computing power demand device, the service request side can, according to the available computing node information, distribute the specific computing task to the corresponding available computing node, and the computing node executes the specific computing task and feeds back the task result of the computing task to the service request side, so as to realize the reasonable allocation of the computing power resources of the service request side. Exemplarily, when the service request side issues a computing task to an available computing node, it can carry the reliability requirement parameter and energy consumption requirement parameter corresponding to the computing task.

[0129] In one embodiment, as shown in Figure 3 Step S31: The service request side 301, as a computing power demand device, can initiate a computing task request to the scheduling server side and send the specific information of the computing power task to the computing power scheduling server side; Step S32: The scheduling server side requests the computing node information from the storage device 303 according to the computing task request; Step S33: The computing node information feeds back a node information list to the scheduling server side; Step S34: The scheduling server side performs computing power scheduling according to the node information list to determine the available computing node information; Step S35: The scheduling server side feeds back a computing task request response to the service request side 301 and feeds back the available computing node information to the service request side 301; Step S36: The service request side issues a computing task to the available computing node; Step S37: The computing node 304 feeds back the task result of the computing task to the service request side 301.

[0130] By carrying the performance requirement parameters of the computing power resources for the current task in the computing task request initiated by the service request side, including the node reliability requirement parameters and node energy consumption requirement parameters of the computing power nodes, it is possible to enable the scheduling server to effectively consider the reliability and energy-saving requirements of the computing power nodes during the computing power scheduling and orchestration process when scheduling the computing power resources. Thus, it is possible to provide the service request side with computing power resources that more conform to the actual task requirements to execute the computing task and provide a more reasonable computing power resource scheduling result.

[0131] In one embodiment, the performance requirement parameters of the computing power resources include: node reliability requirement parameters and node energy consumption requirement parameters.

[0132] Among them, the node reliability requirement parameters can be used to describe the performance requirements of the service request side for the reliability of the computing power node to execute the computing task for the current computing task; the node energy consumption requirement parameters can be used to describe the specific performance requirements for the power energy consumption of the node for the current computing task.

[0133] In one embodiment, the node reliability requirement parameters include at least one of: transmission capacity requirement parameters, task priority parameters, and node task completion rate parameters.

[0134] Specifically, the transmission capacity requirement parameters can include the connection method, such as wired or wireless, and the number of forwarding nodes. For example, according to the importance of the current computing task, the transmission method of the computing task can be pre-configured as wired or wireless. In addition, the transmission capacity requirement parameters can also include the authentication method supported by the computing power node, as well as the supported security algorithm level, complexity, or the specific algorithm type supported.

[0135] The task priority parameter can be used to define the priority of the current computing task among all the tasks executed by the computing power node. For example, in a computing power node shared by multiple tasks, tasks such as communication, sensing, computing power, and AI data computing can be executed. For different tasks, the computing power node can customize the corresponding task priorities; for example, configure the communication task as the highest priority, the sensing task as the second priority, and the computing power task and AI data computing task as the third priority. Or, for dedicated computing power nodes, the priorities of different types of tasks can also be predefined; for example, different priorities are pre-configured for image data computing tasks, traffic data computing tasks, etc. Exemplarily, for the service request side, the current computing power task can be pre-configured to be executed by the computing power node with the highest priority; that is, it is required that the task level of the computing power task in the computing power node is the highest, and a computing power node of this type is selected for processing.

[0136] The node task completion rate parameter can be used to require the task completion probability of computing nodes. For example, the service request side can pre-configure the node task completion rate parameter required for the current computing task to be greater than 90%; that is, it is required that the computing node has a task completion probability of greater than 90% for the current computing task.

[0137] By pre-configuring parameters such as the transmission capacity requirement parameter, task priority parameter, and node task completion rate parameter of the computing node, the performance conditions for screening computing nodes can be limited, thereby improving the reliability of computing task completion.

[0138] In one embodiment, the node energy consumption requirement parameter includes at least one of: node status requirement information, node device type requirement information, node energy consumption type requirement information, and node location requirement information.

[0139] Specifically, the node status requirement information can limit the current working status of the computing node and / or the status when executing a computing task; for example, the node status can be types such as powered on, sleeping, energy-saving, or powered off, abnormal, faulty, etc. The node device type requirement information can be used to limit whether the node device is a device using green energy, such as a device using solar energy or wind energy for power generation; or, whether the energy price in the area where the node device is located belongs to a low-price area. The node energy consumption requirement information can include the power-limited state of the node device; for example, it can be required that the computing node is a power-unlimited device; or, it can also be a device such as a power-limited mobile terminal or mobile base station that supports power limitation. In addition, the node energy consumption type requirement information can also limit the power-saving requirement of the node device; for example, for the computing node of a mobile device, it is required that the power is sufficient, and a node device with insufficient power is not used. In addition, the node energy consumption type requirement information can also require the current remaining power of the node device; for example, a mobile device requires the remaining power to be not less than 80% to ensure that the node device can complete the computing task due to insufficient energy. The node location requirement information can be used to require the environmental temperature of the location where the node device is located; when the environmental temperature of the location where the node device is located is relatively low, the energy consumption for refrigeration / heating can be saved. In addition, the actual distance between the node device and the task request side can also be limited to limit the data transmission delay.

[0140] In one embodiment, according to the target trigger event, the update of the computing node information in the storage device can be triggered. Specifically, refer to Figure 5 As shown, the method further includes:

[0141] Step S51, in response to the target trigger event, send a node information update request to the computing node;

[0142] Step S52: Receive the current node information fed back by the computing power node and save it to the storage device for the storage device to update the computing power node information according to the current node information;

[0143] Step S53: Feed back update response information to the computing power node.

[0144] In one embodiment, the target trigger event includes: a trigger event corresponding to a node periodic update task, or a trigger event corresponding to the computing task request.

[0145] Specifically, for the scheduling server, it can configure a timing task for the node information in the storage device to implement periodic update of the node information. Exemplarily, a timing task can be established in the scheduling server. When the timing task is executed, the scheduling server sends a node information update request to each computing power node. For the computing power node, after receiving the computing power node information update request, it can respond to the request, collect the current node information of the computing power node, feed it back to the scheduling server, and save it to the storage device; the storage device updates the node information according to the current node information. Alternatively, the computing power node can directly send the current node information to the storage device, and the storage device updates the stored node information according to the current node information; thus realizing the periodic update of the node information in the storage device.

[0146] Exemplarily, for the scheduling server, it can also be to trigger the update of the computing power node information according to the computing task request of the service request side. Specifically, after receiving the computing task request from the service request side, the scheduling server can respond to the computing task request and trigger the sending of a node information update request to the computing power node. For the computing power node, after receiving the computing power node information update request, it can respond to the request, collect the current node information of the computing power node, feed it back to the scheduling server, and save it to the storage device; thus realizing the immediate update of the node information in the storage device. By triggering the update of the computing power node information in response to the computing task request of the service request side, when the scheduling server performs computing power resource scheduling for the current computing task request, it can obtain the latest node information, thereby providing a more accurate computing power resource scheduling result.

[0147] Exemplarily, for a storage device, when updating the stored node information, the updated node information may include node energy consumption parameters of computing power nodes, basic node device information, etc. For example, the current idle computing power of the node device and the available space of the memory can be updated; or, the working state of the processor can be updated, for example, normal or abnormal state. For example, the node task completion rate can be updated. For example, parameters such as the node state corresponding to the node energy consumption parameter, the current remaining power, the ambient temperature, and the power limited state can be updated.

[0148] In one embodiment, in the above step S12, the screening of the computing power nodes in the computing power node information according to the computing task request to determine the available computing power node information includes:

[0149] Taking the task parameters and the computing power resource performance requirement parameters as input parameters and inputting them into a computing power resource analysis model to obtain available computing power node information; wherein, the computing power resource analysis model is used to sort the computing power nodes to determine the computing power node sorting result; and output available computing power node information based on the computing power node sorting result.

[0150] Exemplarily, a computing power resource analysis model can be pre-trained on the scheduling server side, and the computing power resource analysis model can be used to screen and arrange available computing power resources for the current computing task. For example, an initial computing power analysis model based on DNN (Deep Neural Networks) can be pre-constructed, historical data of computing power resource allocation can be collected to construct sample data, and the initial computing power analysis model can be trained using the sample data. For example, the initial computing power analysis model can be trained using a reinforcement learning training framework to obtain a trained computing power resource analysis model. Among them, the collected sample data can include the task information of historical computing tasks and the corresponding computing power resource allocation results. In addition, it can also include the reliability parameters of computing power nodes and node energy consumption parameters as described in the above embodiments. Among them, the historical task information can include: task identifier, task type of the computing task, task computing volume, time delay, jitter, task completion result, etc. The computing power resource allocation result can include the processor usage information, storage usage information, energy consumption information, task priority information, network status information of the computing power node executing the computing task, and the computing power resource list corresponding to the computing task during computing power resource scheduling, etc. During model training, the collected sample data can be divided into a training data set and a validation data set. The initial computing power analysis model can be iteratively trained using the training data set, with the historical task information and the corresponding computing power resource list as the input parameters of the model, and the corresponding computing power resource allocation result can be output. The model performance of the initial computing power analysis model after each round of training can be verified and updated using the validation data set. Thus, a trained AI model for computing power resource analysis is obtained.

[0151] Exemplarily, on the computing power server side, when using the computing power resource analysis model to allocate computing power resources for the computing task corresponding to the current computing task request, the current computing power node information, task parameters corresponding to the computing task, and computing power resource performance requirement parameters can be used as input parameters. Based on the node reliability requirement parameters and node energy consumption requirement parameters, the computing power nodes can be sorted and screened to determine the available computing power node information.

[0152] Specifically, when the computing task request includes hardware device requirement parameters for the computing power node, for example, when the hardware type of the node device executing the current computing task is limited by the node device type requirement information, the computing power resource analysis model can first screen the computing power nodes in the current computing power node information according to the hardware device requirement parameters to determine the first candidate computing power node set that meets the hardware device requirement parameters. For example, the hardware device requirement parameters can include processor type, number of processor cores, memory type, memory space, and requirements for mobile terminals or non-mobile terminals, etc.

[0153] After determining the first candidate computing power node set, the computing power nodes can be evaluated according to the node reliability requirement parameter and the node energy consumption requirement parameter, the candidate computing power nodes that meet the performance requirements can be sorted, and the available computing power nodes can be determined according to the sorting result.

[0154] For example, the first candidate computing power node set can be secondarily screened according to the node reliability requirement parameter to determine the second candidate computing power node set; then, the computing power nodes that meet the energy consumption requirements in the second candidate computing power node set can be sorted according to the node energy consumption requirement parameter to determine the node sorting result, and the available computing power nodes can be screened according to the sorting result.

[0155] In one embodiment, the available computing power node information includes: a first available computing power node and a second available computing power node. The method further includes: determining the first available computing power node according to the computing power node sorting result; when the reliability parameter of the first available computing power node is lower than the preset reliability standard, determining the second available computing power node according to the computing power node sorting result for performing the computing task by using the first available computing power node and the second available computing power node respectively.

[0156] Exemplarily, the above-mentioned first available computing power node and second available computing power node can be an available computing power node respectively; or, they can also be sets of available computing power nodes respectively, including multiple computing power nodes. Among them, the above-mentioned preset reliability standard can be the task completion rate parameter carried in the computing task request; or, when the computing task request does not include the reliability requirement parameter, the reliability requirement parameter predefined by the system can be used; for example, the reliability requirement parameter can be defined by one or more of the connection method (wired or wireless) and the task completion rate.

[0157] In one embodiment, the method further includes: generating a computing task energy consumption prediction result based on the task parameter by using the computing power resource analysis model, and sorting the computing power nodes in combination with the computing task energy consumption prediction result.

[0158] Specifically, the computing power resource analysis model can estimate the energy consumption of the computing task according to the task parameter carried in the computing task request and the node energy consumption requirement parameter, and output the computing task energy consumption prediction result; and according to the computing task energy consumption prediction result, in combination with the node reliability requirement parameter and the node energy consumption requirement parameter, determine the available node information. For example, when training the computing power resource analysis model, the energy consumption of each historical task can be used as sample data, and the initial computing power analysis model can be iteratively trained so that the trained computing power resource analysis model can be used to output the computing task energy consumption prediction result of the computing task, and determine the available computing power node information with reference to the computing task energy consumption prediction result.

[0159] Exemplarily, after screening candidate computing power nodes according to energy consumption requirements and reliability requirements, the energy consumption for completing the computing task can be predicted for each candidate node, and the candidate nodes can be sorted according to the prediction result, and available computing power nodes can be selected according to the sorting result.

[0160] Exemplarily, when evaluating candidate computing power nodes, computing power nodes without power saving requirements can be preferentially selected.

[0161] In one embodiment, the available computing power node information includes energy consumption mode parameters and processor frequency band parameters corresponding to the available computing power nodes.

[0162] Specifically, when the scheduling server evaluates the computing power resources to determine the available computing power resources, the model can also provide the recommended processor main frequency of the available node device and the energy consumption mode of the node device according to the performance requirements corresponding to the current calculation and the device performance of the available computing power node device. For example, when the current computing task is of the data type of the service flow and the node device can meet the performance requirements by using a lower main frequency, it is recommended that the available computing power node be configured with a low main frequency. Among them, the energy consumption mode of the node device can be a low energy consumption mode to execute the computing task. Thus, the purpose of energy saving of the node device is achieved.

[0163] Exemplarily, for a computing power node, during the execution of a computing task, the node reliability parameters and node energy consumption parameters can be periodically collected; or, after the computing task is completed, the node reliability parameters and node energy consumption parameters are collected and fed back to the scheduling server for optimizing the model and improving the accuracy of computing power resource scheduling.

[0164] In one embodiment, the method may further include: when the computing task corresponding to the computing task request is splittable, splitting the computing task to obtain a computing sub-task queue; and sequentially executing and configuring the available computing power node information corresponding to each sub-task based on the computing sub-task queue.

[0165] Specifically, after receiving the computing task request from the service request side, the computing power server can parse the computing task according to the task parameters, first determine whether the current computing task is splittable; when it is recognized that the computing task is splittable, the computing task can be split into multiple sub-tasks, sorted according to a preset rule, and a sub-task queue is established. The scheduling server sequentially pulls sub-tasks from the sub-task queue, and in response to the trigger of the sub-task, obtains the computing power node information from the storage device, and determines the corresponding available computing power node information according to the computing power node information. Sequentially pull sub-tasks from the sub-task queue, and calculate the corresponding available computing power node information for each sub-task respectively to implement the execution of each sub-task separately.

[0166] In one embodiment, the method further includes: in response to a software configuration request from an available computing power node for the computing task, feeding back software data corresponding to the computing task or compiled software data to the available computing power node for the available computing power node to execute the computing task by using the software data or the compiled software data.

[0167] Specifically, when the task request end publishes a computing task to the corresponding computing power node, if the computing power node identifies that the application program required to execute the computing task is missing, the computing power node can trigger a software configuration request to the computing power server. The specific requirements of the computing power node for the application program can be carried in this request; for example, whether a compiled application program is required, or the software environment of the computing power node. After receiving the software configuration request, the scheduling server can send the storage address of the software program required by the computing power node, software version information, or the compiled application program to the node device, so that the node device can obtain and install the software data and execute the computing task by using the software data.

[0168] For example, for different computing power node devices, since the processors of the node devices may be different and the system environments may also be different, such as using the Windows system or Linux; it may be necessary to compile the software corresponding to the hardware platform required for the execution of the computing power task. If there is no available software on the selected available computing power node device, but the computing power task can be executed by compilation or by using the software version already stored on the storage device; then, before executing the computing power task, the computing power device needs to compile the software to obtain the software required by the hardware platform, or use the available software stored in other locations to execute the computing power task.

[0169] In one embodiment, after determining the available computing power node according to the sorting result of the candidate nodes, the current load of the available computing power node can also be analyzed. If the load of the available computing power node is higher than a preset first load threshold, a computing power node with a current load lower than a second load threshold can be selected according to the sorting result as the available computing power node.

[0170] For example, for the currently selected available computing power node, if the computing power node with a high execution efficiency for the current task has a too high load, a computing power node that supports the computing power task but has a lower execution efficiency and a lower load can be selected for computing power execution. In addition, methods such as online compilation can be used to provide a software version matching the computing power node executing the computing power task.

[0171] In one embodiment, when the computing power node executes the computing task, the node performance of the computing power node during task execution can also be monitored. Refer to Figure 6 As shown, the method further includes:

[0172] Step S61, collect the node performance detection information corresponding to the available computing nodes when executing the computing task;

[0173] Step S62, when it is identified that the node performance detection information does not meet the preset node performance requirements, determine the standby computing nodes according to the computing node sorting result;

[0174] Step S63, send the standby computing nodes to the service request side for the service request side to allocate the computing task to the standby computing nodes, and the standby computing nodes execute the computing task.

[0175] Specifically, when the computing nodes execute the computing task, the scheduling server can send node performance detection request information to these computing nodes; after the computing node device receives the request information, it can trigger the collection of the node performance detection information within the current time window and upload it to the scheduling server; and the scheduling server analyzes the node performance detection information to determine whether the current detection information of the computing node meets the preset performance requirements. Among them, the node performance detection information can include parameters such as the working state of the node device, the computing speed of the computing task, and the current task completion ratio. If the scheduling server senses that an error has occurred in the computing node currently, or the node performance detection information does not meet the preset node performance requirements, it can notify the computing node to stop executing the computing task. At the same time, according to the sorting result of the candidate computing nodes, select new available computing nodes and push them to the service request side to notify the update of the available computing nodes. Or, if the computing power server identifies that there are no available candidate computing nodes currently, it can trigger the scheduling server to also re-obtain the computing node information from the storage device and recalculate the available computing nodes to achieve the update of the available computing nodes.

[0176] Exemplarily, or, when the scheduling server senses that an error has occurred in the computing node currently, or the node performance detection information does not meet the preset node performance requirements, it can also trigger the scheduling server to directly re-obtain the computing node information from the storage device and recalculate the available computing nodes to achieve the update of the available computing nodes; then send the updated available computing nodes to the service request side, and then the service request side reissues the computing task to the updated computing nodes.

[0177] By monitoring the task execution progress and node performance of the computing nodes when the computing nodes execute the computing task, it is possible to timely sense the computing nodes that do not meet the computing resource performance requirement parameters and update the computing nodes in a timely manner, so as to ensure the stable operation of the computing task, ensure that the service request side can obtain accurate task calculation results, and improve the reliability.

[0178] In one embodiment, the method may further include: when it is recognized that there is no available computing power node in the computing power node information, sending a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

[0179] Exemplarily, when screening computing power nodes according to the hardware requirements for computing power node devices in the computing task request, if it is determined that there is no computing power node meeting the hardware requirements currently, the computing power server may trigger sending a computing power resource scheduling request to an external computing power device. Or, it may also be when screening candidate nodes according to node reliability requirement parameters and node energy consumption requirement parameters, when it is determined that there is no candidate meeting the performance requirements, triggering sending a computing power resource scheduling request to an external computing power device. Or, it may also be when determining available computing power nodes according to the sorting result of candidate nodes and performing a current load analysis on the available computing power nodes, if the load analysis result of the computing power nodes determines that there is no available computing power node meeting the node load requirements currently, it may trigger sending a computing power resource scheduling request to an external computing power device. Or, it may also be when the computing power node executes a computing task and triggers an update to the computing power node, if it is determined that there is no available update node currently, it may trigger sending a computing power resource scheduling request to an external computing power device.

[0180] Exemplarily, referring to Figure 7 As shown, after receiving the computing task request from the service request side, the scheduling server may first determine whether the corresponding computing power task is splittable; if it is splittable, it is split into multiple subtasks and written into the task queue, and each subtask is executed in turn. Secondly, it may also determine whether the computing task request includes the device hardware requirements for the computing power node and identify the device hardware of each computing power node in the computing power resource pool. When it is determined that there is no computing power node meeting the device hardware requirements in the computing power resource pool, it may trigger a request for external computing power resources and select appropriate computing power resources from other candidate resource pools. Or, when it is determined that there is a computing power node meeting the device hardware requirements, evaluate the performance of the computing power node to determine whether it can meet the performance requirements of the current computing task for the computing power node. Among them, the performance requirements of the computing task may include task processing time, task priority (urgency), reliability requirements, jitter tolerance, etc. For different performance requirements, the evaluation priority can be configured. For example, configure the priority of reliability requirements to be higher than the priority of energy consumption requirements; that is, first consider the reliability performance requirements of the computing task for the computing power node, and then consider the energy consumption requirements of the computing task. For example, it can be determined separately whether there is a computing power node device meeting the energy consumption requirements and a computing power node device meeting the reliability requirements; if there is, the available computing power node can be determined. Or, if not, determine whether there is available external computing power resources or high-level computing power resources, and send a computing power scheduling request to the external resources.

[0181] In one embodiment, as Figure 8 shown, a computing power resource scheduling method is provided and applied to a service request side. The method may include the following steps:

[0182] Step S81: Send a computing task request to the computing power resource scheduling side; the computing task request includes task parameters and computing power resource performance requirement parameters.

[0183] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0184] Exemplarily, for the service request side, for a computing task to be executed, the computing power resource performance requirement parameters corresponding to the computing task may be predefined, such as any one of reliability requirement parameters and energy consumption requirement parameters. And a computing task request is generated according to the computing power resource performance requirement parameters and the task parameters of the computing task, and is sent to the computing power scheduling service side.

[0185] Exemplarily, the node reliability requirement parameters include at least one of transmission capacity requirement parameters, task priority parameters, and node task completion rate parameters.

[0186] Exemplarily, the node energy consumption requirement parameters include at least one of node status requirement information, node device type requirement information, node energy consumption type requirement information, and node location requirement information.

[0187] Step S82: Receive the available computing power node information sent by the computing power resource scheduling side; the available computing power node information is determined by the computing power resource scheduling side according to the computing task request by screening the computing power node information.

[0188] Specifically, for the scheduling service side, after receiving the computing task request, it can obtain the current computing power resources, that is, the computing power node information, from the storage device; and evaluate the computing power resources to determine the available computing power node information and feedback it to the service request side. After receiving the available computing power node information, the service request side can feedback a reception response information to the scheduling service side.

[0189] Step S83: Allocate the computing task to the computing power node corresponding to the available computing power node information to obtain the computing task execution result fed back by the computing power node.

[0190] Exemplarily, for the service request side, after obtaining the available computing power node information, it can publish the computing power task to the corresponding available computing power node, and the computing power node executes the corresponding computing task and receives the task execution result fed back by the computing power node.

[0191] In one embodiment, the computing task includes a plurality of sub-tasks that can be split; the method further includes: receiving the available computing power node information corresponding to each sub-task fed back by the computing power resource scheduling end.

[0192] Specifically, when the current computing task is splittable, the scheduling server can split the computing task into several sub-tasks and determine the corresponding available computing power nodes for each sub-task. For the service request end, after receiving the available computing power node information of the sub-task, it can publish the split sub-task to the corresponding available computing power result to obtain the task execution result of the available computing power node for this sub-task. And it can merge according to the execution results of multiple sub-tasks to obtain the complete task execution result corresponding to this computing task.

[0193] In one embodiment, the available computing power node information includes: the first available computing power node, the second available computing power node. The method further includes: allocating the computing task to the first available computing power node and the second available computing power node to respectively obtain the computing task execution results fed back by the first available computing power node and the second available computing power node when executing the computing task.

[0194] Exemplarily, when the reliability of the available computing power node is relatively low, multiple computing power nodes can be used to execute the computing task simultaneously. For the service request end, it can receive the computing task execution results of multiple computing power nodes for the same computing task. If the execution results of multiple computing tasks are the same, it can be confirmed that the computing task execution result is accurate; or, if the execution results of multiple computing tasks are different, the computing task request for this computing task can be re-triggered to re-obtain the computing task execution result.

[0195] In one embodiment, the method further includes: collecting the node performance detection information corresponding to the available computing power node when executing the computing task; when identifying that the node performance detection information does not meet the preset node performance requirements, sending node exception information to the computing power resource scheduling end; obtaining the standby computing power node information sent by the computing power resource scheduling end, and allocating the computing task to the standby computing power node.

[0196] Specifically, for the service request side, when issuing a computing task to a computing power node and the computing power node executes the computing task, it can also send a node performance detection information request to the computing power node to obtain the device performance and status of the computing power node when executing the computing task; and analyze the device performance and status of the node device currently executing the computing task based on the node performance detection information feedback by the computing power node to determine whether there is an abnormality in the node device and whether the execution efficiency of the current computing task can meet the requirements for the task completion time. If it is determined that there is an abnormality in the node device, a node update request can be triggered to be sent to the scheduling server, and the node abnormality information of the abnormal computing power node, such as node performance detection information, can be carried. The scheduling server can reschedule the computing power node according to the node update request and feedback the updated node information to the service server. The service server can send a notice to the original computing power node to stop executing the task and issue the computing task to the updated computing power node to re-execute the computing task. Thus, the smooth completion of the computing task is ensured.

[0197] In one embodiment, the available computing power node information is determined by the computing power resource scheduling side according to the computing task request by screening the computing power node information, including:

[0198] The computing power resource scheduling side uses the task parameters and the computing power resource performance requirement parameters as input parameters and inputs them into the computing power resource analysis model to obtain the available computing power node information;

[0199] Among them, the computing power resource analysis model is used to sort the computing power nodes to determine the computing power node sorting result; and output the available computing power node information based on the computing power node sorting result.

[0200] In one embodiment, the method further includes: the computing power resource scheduling side determines the first available computing power node according to the computing power node sorting result; and when identifying that the reliability parameter of the first available computing power node is lower than the preset reliability standard, determines the second available computing power node according to the computing power node sorting result.

[0201] In one embodiment, the computing task request includes: device hardware requirement parameters. The method further includes: the computing power resource scheduling side identifies the computing power node information according to the device hardware requirement parameters, and when identifying that there is a computing power node that meets the device hardware requirement parameters, calls the computing power resource analysis model to determine the available computing power node information; or, when identifying that there is no computing power node that meets the device hardware requirement parameters, sends a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

[0202] Specifically, for the service request side, the device hardware information of the computing power nodes for executing computing tasks can be pre-configured, such as the processor type, available space, memory level, available space, and so on. To ensure that the computing tasks can be completed within the required time.

[0203] In the computing power resource scheduling method provided in the embodiments of the present application, the service request side can pre-configure corresponding node reliability requirements, node energy consumption requirements, and node device type and other requirement parameters for the computing tasks, so that the scheduling server side can allocate computing power resources that meet the service requirements. On the scheduling server side, by introducing the reliability requirements and energy consumption and energy-saving requirements of the current computing task for the computing power nodes in the computing power orchestration; for example, the reliability requirements can include: transmission reliability, priority of computing power tasks, security, success rate of completing tasks, etc.; the energy consumption and energy-saving requirements can include: energy consumption for completing the computing task, whether it is a mobile terminal, whether there is an energy-saving requirement, etc.; and by matching the corresponding computing power resources for the computing tasks through the above parameters, it is possible to achieve the scheduling of computing power resources from the perspectives of energy consumption and reliability requirements, and realize the reasonable scheduling of computing power tasks, improving the accuracy of computing power resource scheduling.

[0204] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0205] Based on the same inventive concept, the embodiments of the present application also provide a computing power resource scheduling device for implementing the above-mentioned computing power resource scheduling method. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the following computing power resource scheduling device can refer to the limitations on the computing power resource scheduling method in the above text, and will not be repeated here.

[0206] In one embodiment, as Figure 9 shown, a computing power resource scheduling device is provided, which can be applied to the scheduling server side. The device 900 includes:

[0207] A computing task request response module 901, configured to obtain computing power node information from a storage device in response to a computing task request from a service request end; wherein, the computing task request includes: task parameters and computing power resource performance requirement parameters.

[0208] A computing power node determination module 902, configured to screen the computing power node information according to the computing task request to determine available computing power node information.

[0209] A computing task allocation module 903, configured to send the available computing power node information to the service request end and screen the computing power node information according to the computing task request to determine available computing power node information.

[0210] In one embodiment, the apparatus 900 further includes: a node information update module, configured to send a node information update request to a computing power node in response to a target trigger event to obtain current node information fed back by the computing power node, and update the computing power node information of the storage device based on the current node information.

[0211] In one embodiment, the target trigger event includes: a trigger event corresponding to a node periodic update task, or a trigger event corresponding to the computing task request.

[0212] In one embodiment, the apparatus further includes: a software configuration request response module, configured to, in response to a software configuration request of an available computing power node for the computing task, feed back software data corresponding to the computing task or compiled software data to the available computing power node for the available computing power node to execute the computing task using the software data or the compiled software data.

[0213] In one embodiment, the apparatus 900 further includes: a node performance monitoring module, configured to collect node performance detection information corresponding to the available computing power node when executing the computing task; and determine a standby computing power node according to the computing power node sorting result when identifying that the node performance detection information does not meet the preset node performance requirements, so as to execute the computing task using the standby computing power node.

[0214] In one embodiment, the apparatus 900 further includes: an external computing power resource scheduling module, configured to send a computing power resource scheduling request to an external computing power device when identifying that there is no available computing power node in the computing power node information, so as to obtain the allocated external computing power device as the available computing power node information.

[0215] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0216] In one embodiment, asFigure 10 As shown, a computing power resource scheduling device is provided, which can be applied to the service request side. The device 100 includes:

[0217] A request processing module 1001, configured to send a computing task request to the computing power resource scheduling side; the computing task request includes task parameters and computing power resource performance requirement parameters;

[0218] A node information receiving module 1002, configured to receive the available computing power node information sent by the computing power resource scheduling side; the available computing power node information is determined by the computing power resource scheduling side according to the computing task request and screening the computing power node information;

[0219] A computing task allocation module 1003, configured to allocate the computing task to the computing power node corresponding to the available computing power node information, so as to obtain the computing task execution result fed back by the computing power node.

[0220] In one embodiment, the available computing power node information includes: a first available computing power node and a second available computing power node; the computing task allocation module 1003 further includes: a task reallocation module, configured to allocate the computing task to the first available computing power node and the second available computing power node, so as to respectively obtain the computing task execution results fed back by the first available computing power node and the second available computing power node when executing the computing task.

[0221] In one embodiment, the device 100 further includes: a computing power node reset module, configured to collect the node performance detection information corresponding to the available computing power node when executing the computing task; when identifying that the node performance detection information does not meet the preset node performance requirements, send node exception information to the computing power resource scheduling side; so as to obtain the standby computing power node information sent by the computing power resource scheduling side, and allocate the computing task to the standby computing power node.

[0222] In one embodiment, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

[0223] Each module in the above computing power resource scheduling device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in hardware form or independent of it, or stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0224] In one embodiment, an electronic device is provided. The electronic device can be a terminal device such as a mobile phone, a camera, an action camera, a drone, etc. with functions of taking pictures and videos. Its internal structure diagram can be as Figure 6As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. The computer program, when executed by the processor, implements a method for analyzing medical images. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0225] Those skilled in the art can understand that Figure 11 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0226] In one embodiment, an electronic device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0227] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0228] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0229] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0230] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., and are not limited thereto. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., and are not limited thereto.

[0231] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0232] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A computing power resource scheduling method, characterized in that, the method includes: In response to a computing task request from a service request end, obtaining computing power node information from a storage device; wherein, the computing task request includes: task parameters and computing power resource performance requirement parameters; Filter the computing power node information according to the computing task request to determine available computing power node information; Send the available computing power node information to the service request end.

2. The method according to claim 1, characterized in that, the method further includes: In response to a target trigger event, sending a node information update request to a computing power node; Receiving the current node information fed back by the computing power node and saving it to the storage device for the storage device to update the computing power node information according to the current node information; Sending update response information to the computing power node.

3. The method according to claim 2, characterized in that, the target trigger event includes: a trigger event corresponding to a node periodic update task, or a trigger event corresponding to the computing task request.

4. The method according to claim 1, characterized in that, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

5. The method according to claim 1 or 4, characterized in that, filtering the computing power nodes in the computing power node information according to the computing task request to determine available computing power node information includes: Using the task parameters and the computing power resource performance requirement parameters as input parameters and inputting them into a computing power resource analysis model to obtain available computing power node information; wherein, the computing power resource analysis model is used to sort the computing power nodes to determine the computing power node sorting result; and output available computing power node information based on the computing power node sorting result.

6. The method according to claim 5, characterized in that, the method further includes: Using the computing power resource analysis model to generate a computing task energy consumption prediction result based on the task parameters, and sorting the computing power nodes in combination with the computing task energy consumption prediction result.

7. The method according to claim 5, characterized in that, the available computing power node information includes: a first available computing power node and a second available computing power node; the method further includes: Determining a first available computing power node according to the computing power node sorting result; When the reliability parameter of the first available computing power node is lower than a preset reliability standard, determining the second available computing power node according to the computing power node sorting result for using the first available computing power node and the second available computing power node to execute the computing task respectively.

8. The method according to claim 4, characterized in that, the node reliability requirement parameters include at least one of: transmission capacity requirement parameters, task priority parameters, and node task completion rate parameters.

9. The method according to claim 4, characterized in that, the node energy consumption requirement parameters include at least one of: node status requirement information, node device type requirement information, node energy consumption type requirement information, and node location requirement information.

10. The method according to claim 1, characterized in that, The available computing power node information includes the energy consumption mode parameters and processor frequency band parameters corresponding to the available computing power nodes.

11. The method according to claim 1, wherein, the method further includes: In response to a software configuration request from an available computing power node for the computing task, feedback the software data or compiled software data corresponding to the computing task to the available computing power node for the available computing power node to execute the computing task using the software data or compiled software data.

12. The method according to claim 1, wherein, the method further includes: Collect the node performance detection information corresponding to the available computing power node when executing the computing task; When it is identified that the node performance detection information does not meet the preset node performance requirements, determine the standby computing power node according to the computing power node sorting result; Send the standby computing power node to the service request end for the service request end to allocate the computing task to the standby computing power node and for the standby computing power node to execute the computing task.

13. The method according to claim 1, wherein, the method further includes: When it is identified that there is no available computing power node in the computing power node information, send a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

14. The method according to claim 1, wherein, the computing task request includes: device hardware requirement parameters; the method further includes: Identify the computing power node information according to the device hardware requirement parameters, and when it is identified that there is a computing power node that meets the device hardware requirement parameters, call a computing power resource analysis model to determine the available computing power node information; or When it is identified that there is no computing power node that meets the device hardware requirement parameters, send a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

15. The method according to claim 1, wherein, the method further includes: When the computing task corresponding to the computing task request is splittable, perform a splitting process on the computing task to obtain a computing subtask queue; Based on the computing subtask queue, sequentially execute and configure the available computing power node information corresponding to each subtask.

16. The method according to claim 1, wherein, the method further includes: Receive a node registration request of a computing power node; wherein, the node registration request includes the node performance information corresponding to the computing power node; the node performance information includes a node reliability parameter and a node energy consumption parameter; Save the computing power node to a storage device and feedback a registration response message to the computing power node.

17. A computing power resource scheduling method, wherein, the method includes: Send a computing task request to a computing power resource scheduling end; the computing task request includes task parameters and computing power resource performance requirement parameters; Receive the available computing power node information sent by the computing power resource scheduling end; the available computing power node information is determined by the computing power resource scheduling end according to the computing task request by screening the computing power node information. Allocate the computing task to the computing power nodes corresponding to the available computing power node information for obtaining the execution results of the computing task fed back by the computing power nodes.

18. The method according to claim 17, wherein, the computing power resource performance requirement parameters include: node reliability requirement parameters and node energy consumption requirement parameters.

19. The method according to claim 18, wherein, the node reliability requirement parameters include at least one of: transmission capacity requirement parameters, task priority parameters, and node task completion rate parameters.

20. The method according to claim 18, wherein, the node energy consumption requirement parameters include at least one of: node status requirement information, node device type requirement information, node energy consumption type requirement information, and node location requirement information.

21. The method according to claim 17 or 18, wherein, the available computing power node information is determined by the computing power resource scheduling end through screening the computing power node information according to the computing task request, and includes: the computing power resource scheduling end takes the task parameters and the computing power resource performance requirement parameters as input parameters and inputs them into a computing power resource analysis model to obtain the available computing power node information; wherein, the computing power resource analysis model is used to sort the computing power nodes to determine the sorting result of the computing power nodes; and output the available computing power node information based on the sorting result of the computing power nodes.

22. The method according to claim 17 or 18, wherein, the method further includes: the computing power resource scheduling end uses the computing power resource analysis model to generate a computing task energy consumption prediction result based on the task parameters, and sorts the computing power nodes in combination with the computing task energy consumption prediction result.

23. The method according to claim 17, wherein, the available computing power node information includes: a first available computing power node and a second available computing power node; the method further includes: Allocate the computing task to the first available computing power node and the second available computing power node for obtaining the execution results of the computing task fed back by the first available computing power node and the second available computing power node respectively.

24. The method according to claim 23, wherein, the method further includes: the computing power resource scheduling end determines the first available computing power node according to the sorting result of the computing power nodes; and when identifying that the reliability parameter of the first available computing power node is lower than a preset reliability standard, determines the second available computing power node according to the sorting result of the computing power nodes.

25. The method according to claim 17, wherein, the available computing power node information includes the energy consumption mode parameters and processor frequency band parameters corresponding to the available computing power nodes.

26. The method according to claim 17, wherein, the computing task request includes: device hardware requirement parameters; the method further includes: The computing power resource scheduling end identifies the computing power node information according to the device hardware requirement parameters, and when identifying that there is a computing power node that meets the device hardware requirement parameters, calls a computing power resource analysis model to determine the available computing power node information; or when identifying that there is no computing power node that meets the device hardware requirement parameters, sends a computing power resource scheduling request to an external computing power device to obtain the allocated external computing power device as the available computing power node information.

27. The method according to claim 17,[[]] wherein,[[]] the computing task includes a plurality of sub-tasks that can be split; the method further includes:[[]] receiving the available computing power node information corresponding to each sub-task fed back by the computing power resource scheduling end.

28. The method according to claim 17,[[]] wherein,[[]] the method further includes:[[]] collecting the node performance detection information corresponding to the available computing power node when executing the computing task; when identifying that the node performance detection information does not meet the preset node performance requirements, sending node exception information to the computing power resource scheduling end; obtaining the standby computing power node information sent by the computing power resource scheduling end, and allocating the computing task to the standby computing power node.

29. A computing power resource scheduling device,[[]] wherein,[[]] the device includes:[[]] A computing task request response module, configured to obtain computing power node information from a storage device in response to a computing task request from a service request end; wherein, the computing task request includes: task parameters, computing power resource performance requirement parameters; A computing power node determination module, configured to screen the computing power node information according to the computing task request to determine available computing power node information; A computing task allocation module, configured to send the available computing power node information to the service request end and screen the computing power node information according to the computing task request to determine available computing power node information.

30. The device according to claim 29,[[]] wherein,[[]] the device further includes:[[]] A node information update module, configured to send a node information update request to a computing power node in response to a target trigger event to obtain the current node information fed back by the computing power node, and update the computing power node information in the storage device based on the current node information.

31. The device according to claim 29,[[]] wherein,[[]] the target trigger event includes: a trigger event corresponding to a node periodic update task, or, a trigger event corresponding to the computing task request.

32. The device according to claim 29,[[]] wherein,[[]] the device further includes:[[]] A software configuration request response module, configured to respond to a software configuration request of an available computing power node for the computing task, and feed back software data or compiled software data corresponding to the computing task to the available computing power node, so that the available computing power node uses the software data or compiled software data to execute the computing task.

33. The device according to claim 29,[[]] wherein,[[]] the device further includes:[[]] A node performance monitoring module, configured to collect node performance detection information corresponding to the available computing power nodes when executing the computing task; and determine a standby computing power node according to the computing power node sorting result when identifying that the node performance detection information does not meet the preset node performance requirements, so as to use the standby computing power node to execute the computing task.

34. The apparatus according to claim 29, wherein, the apparatus further includes: An external computing power resource scheduling module, configured to send a computing power resource scheduling request to an external computing power device when identifying that there is no available computing power node in the computing power node information, so as to obtain the allocated external computing power device as the available computing power node information.

35. The apparatus according to claim 29, wherein, The computing power resource performance requirement parameters include: node reliability requirement parameters, node energy consumption requirement parameters.

36. A computing power resource scheduling apparatus, wherein, the apparatus includes: A request processing module, configured to send a computing task request to a computing power resource scheduling end; the computing task request includes task parameters and computing power resource performance requirement parameters; A node information receiving module, configured to receive the available computing power node information sent by the computing power resource scheduling end; the available computing power node information is determined by the computing power resource scheduling end according to the computing task request to screen the computing power node information; A computing task allocation module, configured to allocate the computing task to the computing power node corresponding to the available computing power node information, so as to obtain the computing task execution result fed back by the computing power node.

37. The apparatus according to claim 36, wherein, The available computing power node information includes: a first available computing power node, a second available computing power node; the computing task allocation module further includes: A task reallocation module, configured to allocate the computing task to the first available computing power node and the second available computing power node, so as to respectively obtain the computing task execution results fed back by the first available computing power node and the second available computing power node when executing the computing task.

38. The apparatus according to claim 36, wherein, the apparatus further includes: A computing power node reset module, configured to collect node performance detection information corresponding to the available computing power node when executing the computing task; when identifying that the node performance detection information does not meet the preset node performance requirements, send node exception information to the computing power resource scheduling end; so as to obtain the standby computing power node information sent by the computing power resource scheduling end, and allocate the computing task to the standby computing power node.

39. The apparatus according to claim 36, wherein, The computing power resource performance requirement parameters include: node reliability requirement parameters, node energy consumption requirement parameters.

40. An electronic device, including a memory and a processor, the memory stores a computer program, wherein, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 16 or 17 to 28.

41. A computer-readable storage medium, on which a computer program is stored, wherein, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 16 or claims 17 to 28.

42. A computer program product, characterized in that the computer program product includes a computer program which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 16 or claims 17 to 28.

Images