An edge computing node task scheduling method and device based on thermal coupling constraints and a storage medium

Abstract

This application discloses a method, apparatus, and storage medium for scheduling edge computing nodes based on thermal coupling constraints. The method includes: constructing thermal coupling characterization parameters based on the spatial adjacency relationship between the target edge computing node and adjacent edge computing nodes; extracting task feature parameters from the tasks to be scheduled and generating task thermal load characterization values; mapping the tasks to be scheduled to at least two different thermal shadow segments; constructing corresponding dispatch gating functions for each thermal shadow segment; generating corresponding dispatch tokens for each task to be scheduled according to the dispatch permission level; and mapping each task to be scheduled to a target queue of the target edge computing node based on the dispatch tokens, and controlling the tasks to be scheduled to enter the execution state from the queue according to the release condition of the dispatch gating function.

Description

Technical Field

[0001] This application relates to the field of edge computing technology, and in particular to an edge computing node task scheduling method, apparatus and storage medium based on thermal coupling constraints. Background Technology

[0002] With the development of edge computing technology, more and more computing tasks are being deployed to edge computing nodes closer to the data source to meet business requirements for low latency and high real-time performance. In practical applications, multiple edge computing nodes are usually deployed in a dense manner, and there is a certain spatial thermal impact relationship between different nodes.

[0003] Existing edge computing task scheduling schemes typically allocate tasks based on node load, task priority, or resource occupancy to improve overall task processing efficiency. While some schemes incorporate node temperature monitoring mechanisms, most only adjust the temperature status of a single node, such as migrating tasks, delaying tasks, or adjusting processing frequency when a node's temperature exceeds a threshold.

[0004] However, in scenarios where edge computing nodes are densely deployed, the thermal impact between adjacent nodes can spread during task execution. When multiple high-heat-load tasks are executed in a local area, the thermal impact between nodes can easily overlap, resulting in an aggravated temperature rise in the local area.

[0005] Existing technologies lack an effective understanding of the thermal impact relationships between nodes, making it difficult to stably control different thermal load tasks when they enter the execution process. This can easily lead to large fluctuations in the thermal state of nodes, affecting the operational stability of edge computing systems under thermal constraints. Summary of the Invention

[0006] To address the aforementioned technical issues, this application provides a method, apparatus, and storage medium for scheduling edge computing node tasks based on thermal coupling constraints.

[0007] The technical solution provided in this application is described below:

[0008] The first aspect of this application provides a task scheduling method for edge computing nodes based on thermal coupling constraints, applied in a distributed computing environment containing at least one target edge computing node and at least one adjacent edge computing node, the method comprising: Based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing nodes, thermal coupling characterization parameters are constructed; Extract task feature parameters from the task to be scheduled, and generate task thermal load characterization values ​​based on the expected power consumption intensity in the task feature parameters; Based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter, the task to be scheduled is mapped to at least two different thermal shadow segments, and different thermal shadow segments correspond to different thermal disturbance level ranges. For each of the thermal shadow segments, a corresponding dispatch gating function is constructed. The dispatch gating function generates a dispatch permission degree based on the task thermal load characterization value and the thermal coupling characterization parameter. Based on the dispatch permission level, generate a corresponding dispatch token for each task to be scheduled. Based on the dispatch token, each task to be scheduled is mapped to the target queue of the target edge computing node, and the task to be scheduled is controlled to enter the execution state from the queue according to the release condition of the dispatch gate function.

[0009] Optionally, based on the task thermal load characterization value and the inter-node thermal influence degree characterized by the thermal coupling characterization parameter, the task to be scheduled is mapped to at least two different thermal shadow segments, where different thermal shadow segments correspond to different thermal disturbance level ranges, including: Calculate the node thermal coupling influence factor of the target edge computing node in the current scheduling cycle based on the set of thermal coupling characterization parameters. The task thermal load characterization value is coupled with the node thermal coupling influence factor to obtain a comprehensive thermal load index. This comprehensive thermal load index is used to characterize the degree of comprehensive thermal disturbance caused by the task when it is executed on the target edge computing node. Based on the heat dissipation margin parameters of the target edge computing node and the preset thermal safety constraints, at least two segmented thresholds are generated for the comprehensive thermal load index.

[0010] The comprehensive heat load index is compared with the segmented threshold, and the heat disturbance level range to which the task to be scheduled belongs is determined based on the comparison result; Based on the thermal disturbance level range, assign a corresponding thermal shadow segment identifier to the task to be scheduled.

[0011] Optionally, a corresponding dispatch gating function is constructed for each of the thermal shadow segments. The dispatch gating function generates a dispatch permission degree based on the task thermal load characterization value and the thermal coupling characterization parameter, including: Based on the thermal disturbance level range corresponding to the thermal shadow segment, a corresponding initial gating parameter set is determined, wherein the gating parameter set includes at least a basic permissible threshold and an adjustment coefficient; Extract the degree of thermal influence between the target edge computing node and its adjacent edge computing nodes, and generate a thermal coupling correction factor; The task thermal load characterization value is input into a preset load response mapping relationship to obtain the thermal response amount generated by the task to the target edge computing node; The initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity are coupled to obtain the dispatch gating function; The task thermal load characterization value is substituted into the dispatch gating function for evaluation to obtain the corresponding dispatch permission degree.

[0012] Optionally, the initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity are coupled to obtain a dispatch gating function, including: Based on the initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity, corresponding gating influence components are constructed respectively. Nonlinear modulation is applied to the gated influence component, so that the dispatch gate function exhibits nonlinear suppression characteristics when thermal coupling is enhanced or the task thermal load increases; The gated influence components after nonlinear modulation are combined and operated to generate the dispatch gated function.

[0013] Optionally, generating a corresponding dispatch token for each task to be scheduled based on the dispatch permission includes: Determine the permission level to which the task to be scheduled belongs; Based on the permission level, a corresponding token attribute is determined for the task to be scheduled; The token attributes are modified based on the current thermal environment; The modified token attributes are encoded to generate a dispatch token associated with the task to be scheduled.

[0014] Optionally, the step of mapping each task to be scheduled to a target queue of a target edge computing node based on the dispatch token, and controlling the task to be scheduled to enter the execution state from the queue according to the release condition of the dispatch gating function, includes: The execution path information in the dispatch token is parsed, and the target queue type corresponding to the task to be scheduled is determined based on the execution path information. The target queue type includes at least an immediate execution queue and a delayed execution queue. Write the task to be scheduled into the target queue according to the admission constraint information in the dispatch token; Within the preset scheduling period, the corresponding dispatch gating function is called one by one for the scheduled tasks in the target queue, and the release condition is determined based on the admission constraint information. For scheduled tasks that meet the release conditions, priority is sorted according to the token attributes in the dispatch token; Based on the priority sorting result, the scheduled tasks that meet the release conditions are removed from the target queue and the scheduled tasks are controlled to enter the execution state; Tasks that have not met the release conditions are kept in the target queue.

[0015] Optionally, the thermal shadow segment is a thermal disturbance mapping interval divided based on the task thermal load characterization value and the degree of thermal influence between nodes.

[0016] Optionally, the task thermal load characterization value is obtained based on the mapping relationship between the task's computational load estimate and power consumption intensity.

[0017] A second aspect of this application provides an edge computing node task scheduling device based on thermal coupling constraints, applied in a distributed computing environment comprising at least one target edge computing node and at least one adjacent edge computing node, the device comprising: The parameter construction unit is used to construct thermally coupled characterization parameters based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing node; The load characterization value generation unit is used to extract task feature parameters from the task to be scheduled, and generate task thermal load characterization values ​​based on the expected power consumption intensity in the task feature parameters. The thermal shadow segment mapping unit is used to map the task to be scheduled to at least two different thermal shadow segments based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter. Different thermal shadow segments correspond to different thermal disturbance level ranges. The function construction unit is used to construct corresponding dispatch gating functions for the thermal shadow segments respectively. The dispatch gating functions generate dispatch permission based on the task thermal load characterization value and the thermal coupling characterization parameter. The token generation unit is used to generate a corresponding distribution token for each task to be scheduled based on the distribution permission level. The queue control unit is used to map each scheduled task to a target queue of the target edge computing node based on the dispatch token, and to control the scheduled task to enter the execution state from the queue according to the release condition of the dispatch gate function.

[0018] A third aspect of this application provides an edge computing node task scheduling device based on thermal coupling constraints, the device comprising: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to execute the first aspect and any one of the optional methods in the first aspect.

[0019] A fourth aspect of this application provides a computer-readable storage medium on which a program is stored, which, when executed on a computer, performs the methods of the first aspect and any one of the first aspects.

[0020] As can be seen from the above technical solutions, this application has the following beneficial effects: 1. This application provides an edge computing node task scheduling method based on thermal coupling constraints. By constructing thermal coupling characterization parameters between the target edge computing node and adjacent edge computing nodes, and combining the task thermal load characterization value of the task to be scheduled, the task to be scheduled is divided into thermal shadow segments, so that tasks with different thermal disturbance levels enter the corresponding thermal shadow segments. Furthermore, corresponding dispatch gating functions are constructed for different thermal shadow segments, and dispatch permission and corresponding dispatch tokens are generated based on the dispatch gating functions, thereby controlling the process of the task to be scheduled entering the target queue and entering the execution state.

[0021] 2. Compared with the existing technology that only schedules tasks based on the local temperature or resource occupancy of nodes, this invention does not passively adjust the thermal state of nodes after the task enters execution. Instead, it introduces the thermal influence relationship between nodes and the thermal load characteristics of the task into the task execution admission process. Through the coordinated cooperation between thermal shadow segments, dispatch gating functions and dispatch tokens, the task to be scheduled can be hierarchically controlled according to the current thermal influence state before entering the execution state.

[0022] 3. This application can reduce the superposition of thermal effects caused by high heat load tasks entering the execution state in a local area, mitigate the temperature rise fluctuation in the local area of ​​the edge computing node, and improve the stability of the task scheduling process under thermal constraints. At the same time, by dynamically controlling the process of task entry into the execution state through dispatching gating functions and dispatching tokens, it can also improve the adaptability between the task execution process and the node thermal state. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic flowchart of an embodiment of the edge computing node task scheduling method based on thermal coupling constraints provided in this application. Figure 2 This is a schematic flowchart of an embodiment of step S103 in the edge computing node task scheduling method based on thermal coupling constraints provided in this application. Figure 3 This is a schematic flowchart of an embodiment of step S104 in the edge computing node task scheduling method based on thermal coupling constraints provided in this application. Figure 4 This is a schematic diagram of an embodiment of the edge computing node task scheduling device based on thermal coupling constraints provided in this application. Figure 5 This is a schematic diagram of an embodiment of another edge computing node task scheduling device based on thermal coupling constraints provided in this application. Detailed Implementation

[0025] In this application, the execution subject of the edge computing node task scheduling method based on thermal coupling constraints can be any computing device or computing system with data processing capabilities and task scheduling capabilities, including but not limited to edge computing nodes, edge servers, scheduling nodes in edge computing clusters, cloud-edge collaborative scheduling platforms, distributed control devices, or other processing devices capable of executing task scheduling control processes.

[0026] In some embodiments, the execution entity can be independently deployed in a single edge computing node, and the corresponding edge computing node performs hot-constraint scheduling control on the tasks to be scheduled within this node; in other embodiments, the execution entity can also be deployed in a collaborative scheduling node among multiple edge computing nodes to perform unified scheduling control on the task execution process among multiple edge computing nodes.

[0027] It should be noted that the relationship between the target edge computing node and the adjacent edge computing node described in this application is not limited to physical adjacency, but can also be a node relationship with thermal influence propagation, heat dissipation correlation or thermal coupling influence.

[0028] Furthermore, the tasks described in this application may be data processing tasks, inference tasks, computing tasks, transcoding tasks, cache processing tasks, or other task types that can cause changes in the thermal state of edge computing nodes, and this invention does not limit them.

[0029] Those skilled in the art will understand that, without departing from the core ideas of this invention, the deployment form, system architecture form, and functional implementation method of the above-mentioned execution entity can be adjusted or replaced without affecting the implementation of this invention.

[0030] Please see Figure 1This application first provides an embodiment of an edge computing node task scheduling method based on thermal coupling constraints. In this embodiment, the edge computing cluster includes at least one target edge computing node and at least one adjacent edge computing node. The target edge computing node can be the node currently executing task scheduling control, and the adjacent edge computing nodes can be other edge computing nodes that have a thermal influence propagation relationship with the target edge computing node.

[0031] It should be noted that the adjacent edge computing nodes in this embodiment are not limited to nodes that are physically adjacent, but may also include nodes that can have a thermal impact on the target edge computing node in the heat dissipation path, airflow path or device layout structure.

[0032] This embodiment includes: S101. Based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing node, construct thermal coupling characterization parameters; In this embodiment, the target edge computing node can periodically acquire its own thermal state information and that of its neighboring edge computing nodes.

[0033] The thermal status information may include the node's current temperature; the node's historical temperature trend; the node's power consumption status; the node's heat dissipation capacity status; the node's fan speed; and the ambient temperature inside the node's chassis.

[0034] Furthermore, the target edge computing nodes can establish thermal impact relationships between nodes based on deployment distance, spatial layout, and heat propagation direction. For example, in an edge server rack, two adjacent edge computing nodes may share a portion of the heat dissipation airflow path. When one node continuously performs a high-power task, the heat generated may cause a local temperature increase in the other node.

[0035] Based on the aforementioned inter-node thermal influence relationships, thermal coupling characterization parameters can be constructed to describe the degree of thermal influence between nodes. In some implementations, these thermal coupling characterization parameters can be represented by thermal coupling coefficients, thermal influence matrices, heat propagation weights, or other data formats that can reflect the degree of thermal influence between nodes.

[0036] S102. Extract task feature parameters from the task to be scheduled, and generate a task thermal load characterization value based on the expected power consumption intensity in the task feature parameters. In this embodiment, once a task to be scheduled enters the task scheduling process, task characteristic analysis can be performed first. The task characteristic parameters may include: task computational load; predicted CPU utilization; GPU call intensity; memory access frequency; data read / write frequency; and estimated execution time.

[0037] Furthermore, based on the resource usage and historical execution data of the task to be scheduled, the potential power consumption level during task execution can be predicted to obtain the expected power intensity of the task. For example, for video inference tasks, since they require continuous GPU resource usage during execution, their expected power intensity can be higher than that of ordinary data caching tasks. After obtaining the expected power intensity, a corresponding task thermal load characterization value can be generated. This task thermal load characterization value can be used to represent the degree of impact that the task to be scheduled may have on the node's thermal state during execution.

[0038] S103. Based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter, the task to be scheduled is mapped to at least two different thermal shadow segments, and different thermal shadow segments correspond to different thermal disturbance level ranges. In this embodiment, the thermal disturbance level of the task to be scheduled can be divided by combining the task's thermal load characterization value and the degree of thermal impact currently experienced by the target edge computing node. The thermal shadow segment can be understood as a task mapping interval established for different thermal disturbance levels. For example, the thermal shadow segment can be divided into a low thermal disturbance segment, a medium thermal disturbance segment, and a high thermal disturbance segment. Different thermal shadow segments correspond to different thermal disturbance level intervals. For example, when the thermal load characterization value of a task to be scheduled is high, and the current target edge computing node is strongly affected by the thermal impact of adjacent edge computing nodes, the task to be scheduled can be mapped to the high thermal disturbance segment. Tasks with lower thermal loads can be mapped to the low thermal disturbance segment.

[0039] In one optional embodiment, one implementation of step S103 includes: 1031. Calculate the node thermal coupling influence factor of the target edge computing node in the current scheduling cycle based on the set of thermal coupling characterization parameters; In this embodiment, the target edge computing node can obtain the corresponding set of thermal coupling characterization parameters at the beginning of the current scheduling period, and calculate the corresponding node thermal coupling influence factor within the current scheduling period based on the set of thermal coupling characterization parameters.

[0040] The thermal coupling characterization parameter set can include thermal influence parameters between the target edge computing node and each adjacent edge computing node, such as node distance parameters, heat propagation weight parameters, node temperature difference parameters, and heat diffusion direction parameters.

[0041] Furthermore, the target edge computing node can combine the operating status of each adjacent edge computing node within the current scheduling cycle to statistically analyze the degree of thermal impact corresponding to different adjacent edge computing nodes, thereby forming a comprehensive thermal impact result for the current scheduling cycle. For example, when multiple adjacent edge computing nodes simultaneously execute high-power tasks, the degree of external thermal impact on the target edge computing node may significantly increase due to the continuous accumulation of heat in local areas, and the corresponding node thermal coupling impact factor can be increased accordingly. Conversely, when the overall load of adjacent edge computing nodes is low or some nodes are idle, the degree of thermal impact around the target edge computing node is relatively weak, and the corresponding node thermal coupling impact factor can be reduced accordingly. In some implementations, the node thermal coupling impact factor can be represented by scalar parameters, vector parameters, or matrix parameters, and this application does not limit this.

[0042] 1032. The task thermal load characterization value is coupled with the node thermal coupling influence factor to obtain a comprehensive thermal load index, which is used to characterize the degree of comprehensive thermal disturbance caused by the task when it is executed on the target edge computing node. In this embodiment, after obtaining the node thermal coupling influence factor, the task thermal load characterization value corresponding to the task to be scheduled can be further coupled with the node thermal coupling influence factor to obtain the corresponding comprehensive thermal load index. The task thermal load characterization value can be used to characterize the degree of thermal load that the task itself may generate during execution, while the node thermal coupling influence factor is used to characterize the degree of external thermal influence in the environment where the target edge computing node is currently located. Furthermore, the comprehensive thermal load index can be used to reflect the comprehensive thermal disturbance results that the task to be scheduled may generate when it is executed in the current thermal environment. For example, for the same type of task to be scheduled, when the thermal influence of the area around the target edge computing node is low, its corresponding comprehensive thermal load index can be relatively low; while when there is already a high thermal accumulation state in the area around the target edge computing node, even if the thermal load of the task itself has not changed, its corresponding comprehensive thermal load index may still increase.

[0043] In some implementations, the coupling calculation process may be processed by correlation mapping, thermal effect superposition, nonlinear fusion or other methods to reflect the correlation of thermal effects, and this application does not limit this.

[0044] 1033. Based on the heat dissipation margin parameters of the target edge computing node and the preset thermal safety constraints, generate at least two segmented thresholds for the comprehensive thermal load index.

[0045] In this embodiment, after obtaining the comprehensive thermal load index, the corresponding segmented thresholds can be generated by combining the current heat dissipation margin parameter of the target edge computing node and the preset thermal safety constraints. The heat dissipation margin parameter can be used to characterize the remaining heat dissipation capacity of the target edge computing node. For example, when the overall temperature of the target edge computing node is low and the heat dissipation device is in a high-efficiency state, its corresponding heat dissipation margin parameter can be high; while when the target edge computing node is already in a continuous high-temperature operation state, its corresponding heat dissipation margin parameter may decrease. The preset thermal safety constraints can be used to limit the range of thermal states that the target edge computing node is allowed to reach.

[0046] 1034. Compare the comprehensive heat load index with the segmented threshold, and determine the heat disturbance level range to which the task to be scheduled belongs based on the comparison result; In this embodiment, the comprehensive heat load index corresponding to the task to be scheduled can be compared with the aforementioned segmented thresholds, and the thermal disturbance level range to which the task to be scheduled belongs can be determined based on the comparison result. When the comprehensive heat load index is lower than the first segmented threshold, the corresponding task to be scheduled can be determined to belong to the low thermal disturbance level range; when the comprehensive heat load index is between the first and second segmented thresholds, it can be determined to belong to the medium thermal disturbance level range; and when the comprehensive heat load index is higher than the higher segmented threshold, it can be determined to belong to the high thermal disturbance level range. In some implementations, different thermal disturbance level ranges can correspond to different degrees of thermal impact risk. For example, tasks corresponding to the high thermal disturbance level range may be more likely to cause a rapid rise in node temperature during execution, so a relatively stricter scheduling control method can be adopted subsequently.

[0047] 1035. Assign a corresponding thermal shadow segment identifier to the task to be scheduled based on the thermal disturbance level range.

[0048] In this embodiment, after determining the thermal disturbance level range to which the task to be scheduled belongs, a corresponding thermal shadow segment identifier can be assigned to the task to be scheduled based on the corresponding thermal disturbance level range. In this embodiment, the thermal shadow segment identifier can be used to represent the thermal disturbance level category currently corresponding to the task to be scheduled. Furthermore, different thermal shadow segment identifiers can correspond to different execution admission strategies in subsequent scheduling processes.

[0049] For example, scheduled tasks with a high heat shadow segment identifier can be subject to stricter admission restrictions before entering the execution state, in order to reduce the local heat accumulation phenomenon caused by multiple high heat load tasks entering the execution state at the same time.

[0050] S104. Construct corresponding dispatch gating functions for each of the thermal shadow segments. The dispatch gating functions generate dispatch permission based on the task thermal load characterization value and the thermal coupling characterization parameter. In this embodiment, corresponding dispatch gating functions can be established for different heat shadow segments. These dispatch gating functions describe the admission criteria for tasks to enter the execution state. For example, for tasks in a high-heat disturbance segment, a stricter dispatch gating function can be used to reduce the probability of such tasks directly entering the execution state under high heat conditions. Conversely, for low-heat disturbance tasks, a more lenient admission control condition can be applied. Through this method, the process of a task entering the execution state can be adapted to the current heat state.

[0051] See Figure 3 This application provides an implementation of step S104, including: S1041. Determine the corresponding initial gating parameter set based on the thermal disturbance level range corresponding to the thermal shadow segment. The gating parameter set includes at least a basic permissible threshold and an adjustment coefficient. In this embodiment, after a task to be scheduled is assigned a corresponding thermal shadow segment, an initial gating parameter set can be determined for the task based on the thermal disturbance level range corresponding to that thermal shadow segment. Different thermal disturbance level ranges can correspond to different levels of task execution risk, thus allowing for the configuration of different gating control parameters. For example, a lower thermal disturbance level range can correspond to more lenient execution admission conditions, while a higher thermal disturbance level range can correspond to stricter admission control conditions. Furthermore, the initial gating parameter set can at least include a basic permission threshold and an adjustment coefficient. The basic permission threshold can be used to characterize the basic admission conditions that the task to be scheduled must meet to be allowed to enter the execution state; the adjustment coefficient can be used to dynamically adjust the dispatch gating function during subsequent thermal state changes.

[0052] In some implementations, the base permission thresholds corresponding to different thermal shadow segments can be different. For example, the base permission threshold corresponding to a high thermal disturbance level can be higher than the base permission threshold corresponding to a low thermal disturbance level, in order to reduce the probability that a high thermal load task will directly enter the execution state under high thermal environment.

[0053] S1042. Extract the degree of thermal influence between the target edge computing node and adjacent edge computing nodes, and generate a thermal coupling correction factor. In this embodiment, the degree of thermal impact between the target edge computing node and its adjacent edge computing nodes can be further extracted, and a corresponding thermal coupling correction factor can be generated based on this degree of thermal impact. The degree of thermal impact reflects the influence of adjacent edge computing nodes on the current thermal state of the target edge computing node. For example, when an adjacent edge computing node continuously executes a high-power task, the heat it generates may have an additional thermal impact on the target edge computing node through airflow, rack structure, or heat conduction between devices. After obtaining the degree of thermal impact, a corresponding thermal coupling correction factor can be generated. This thermal coupling correction factor can be used to adjust the task admission relationship in subsequent dispatch gating functions. For example, when the target edge computing node is currently under strong thermal impact, the corresponding thermal coupling correction factor can be increased, thereby strengthening the admission restrictions of subsequent dispatch gating functions for high-heat-load tasks.

[0054] S1043. Input the task thermal load characterization value into a preset load response mapping relationship to obtain the thermal response amount generated by the task to the target edge computing node; In this embodiment, the task thermal load characterization value corresponding to the task to be scheduled can be input into a preset load response mapping relationship to obtain the corresponding thermal response quantity. The load response mapping relationship is used to characterize the correlation between the task thermal load and the thermal state change of the target edge computing node. Through the load response mapping relationship, the task thermal load characterization value can be converted into the corresponding thermal response result to reflect the degree of thermal impact of the task to be scheduled on the target edge computing node during execution.

[0055] Furthermore, the thermal response quantity can be used to describe the thermal state change trend of the task to be scheduled under the current node's operating environment. Different tasks to be scheduled may have different thermal response quantities due to differences in resource allocation methods, execution duration, and computational intensity. In this embodiment, by generating thermal response quantities, the dispatch gating control process can be equipped with the ability to dynamically characterize the actual thermal impact of tasks.

[0056] S1044. Couple the initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity to obtain the dispatch gating function; In this embodiment, after obtaining the initial gating parameter set, thermal coupling correction factor, and thermal response quantity, these parameters can be further coupled to generate the corresponding dispatch gating function. The initial gating parameter set provides the basic control conditions for the task admission process; the thermal coupling correction factor reflects the degree of inter-node thermal influence currently experienced by the target edge computing node; and the thermal response quantity reflects the potential thermal state impact of the scheduled task on the target edge computing node. Furthermore, based on the correlation between the above parameters, the admission conditions corresponding to the task entering the execution state can be dynamically adjusted to form the corresponding dispatch gating function. In this embodiment, the dispatch gating function can be used to describe the permissible changes in the task entering the execution state under the current thermal environment state, so that the task execution process can be adapted to the current thermal state of the target edge computing node.

[0057] S1045. Substitute the task thermal load characterization value into the dispatch gating function to obtain the corresponding dispatch permission degree.

[0058] In this embodiment, after constructing the corresponding dispatch gating function, the task hot load characterization value corresponding to the task to be scheduled can be substituted into the dispatch gating function for evaluation to obtain the corresponding dispatch permission degree. The dispatch permission degree can be used to characterize the degree to which the task to be scheduled is allowed to enter the execution state.

[0059] In this embodiment, the permission assignment degree can be dynamically changed according to changes in task thermal load, inter-node thermal influence status, and the current thermal environment status of the target edge computing node.

[0060] S105. Generate a corresponding dispatch token for each task to be scheduled based on the dispatch permission level. In this embodiment, a corresponding dispatch token can be generated based on the dispatch permission level. The dispatch token can be used to represent the current execution admission status of the task to be scheduled. In some implementations, different dispatch tokens can correspond to different scheduling control strategies. For example: The first type of token distribution allows immediate entry into the execution queue; The second type of token distribution allows for delayed execution; The third type of token distribution corresponds to the waiting period for subsequent hot status release.

[0061] Furthermore, the tokens can also carry control information related to task scheduling, such as target queue information, access restriction information, or scheduling priority information.

[0062] Specifically, in one embodiment, one implementation of this step includes: determining the permission level to which the task to be scheduled belongs; determining a corresponding token attribute for the task to be scheduled based on the permission level; modifying the token attribute according to the current thermal environment; and encoding the modified token attribute to generate a dispatch token associated with the task to be scheduled.

[0063] Specifically, in this embodiment, after obtaining the dispatch permission level corresponding to the task to be scheduled, the current permission level of the task to be scheduled can be determined based on the permission range corresponding to the dispatch permission level. Different permission levels are used to distinguish the degree to which the task to be scheduled is allowed to enter the execution state in the current hot environment.

[0064] In one alternative implementation, the tasks to be scheduled can be divided into multiple permission level ranges according to the degree of permission issued, from high to low. Tasks with higher permission levels indicate that they pose a relatively low risk of thermal disturbance to the target edge computing node under the current thermal state conditions; tasks with lower permission levels indicate that they may cause greater thermal state fluctuations after entering the execution state.

[0065] In this embodiment, the token attribute is used to describe the control characteristics of the task to be scheduled during the subsequent scheduling process. The token attribute may include information such as the target queue type, execution restriction status, admission control status, and scheduling priority status corresponding to the task. The token attributes corresponding to different permission levels are different. For example, the token attribute corresponding to the task to be scheduled at a higher permission level can allow the task to directly enter the immediate execution path; while the token attribute corresponding to the task to be scheduled at a lower permission level can correspond to stricter waiting control conditions or delayed execution conditions.

[0066] After the token attributes are determined, they can be modified based on the current thermal environment status. The current thermal environment status reflects the real-time thermal changes in the area where the target edge computing node is located, including the node's current heat accumulation status, the thermal impact status of adjacent edge computing nodes, and the current heat dissipation status. When the target edge computing node is in a continuous heating phase, the execution restrictions in the token attributes can be strengthened to reduce the probability of high-heat-load tasks entering the execution state; while after the node's thermal state gradually recovers, the corresponding restrictions can be appropriately reduced. Furthermore, the adjusted token attributes can be encoded and converted to form a unified token data structure from multiple scheduling control features. The encoded and generated dispatch token can establish an association with the corresponding task to be scheduled and serve as the basis for scheduling control of the corresponding task in subsequent target queue mapping, task release control, and execution state switching processes.

[0067] This embodiment allows the thermal state change results to be correlated with the task execution control process, enabling the task scheduling process to dynamically adjust the corresponding execution admission state as the node's thermal environment changes.

[0068] S106. Based on the dispatch token, each task to be scheduled is mapped to the target queue of the target edge computing node, and the task to be scheduled is controlled to enter the execution state from the queue according to the release condition of the dispatch gate function.

[0069] In this embodiment, the tasks to be scheduled can be written into the corresponding target queue according to the control information in the dispatch token. For tasks with low current thermal disturbance and that meet the execution conditions, they can be directly written into the immediate execution queue.

[0070] For tasks with high current hot load, they can be placed in a delayed execution queue to wait for subsequent release.

[0071] Furthermore, while the task is in the target queue, the thermal state changes of the current target edge computing node can be continuously monitored.

[0072] When the release condition corresponding to the dispatch gating function is met, the corresponding scheduled task can be released from the target queue to the execution state.

[0073] For example, when the thermal state of the target edge computing node drops to within a preset range, some high-heat-load tasks can be allowed to enter the execution state.

[0074] In one optional embodiment, this step can be implemented as follows: The execution path information in the dispatch token is parsed, and the target queue type corresponding to the task to be scheduled is determined based on the execution path information. The target queue type includes at least an immediate execution queue and a delayed execution queue. The task to be scheduled is written into the target queue according to the admission constraint information in the dispatch token. Within a preset scheduling period, the corresponding dispatch gating function is called one by one for the tasks to be scheduled in the target queue, and it is determined whether the release condition is met based on the admission constraint information. For the tasks to be scheduled that meet the release condition, they are prioritized according to the token attributes in the dispatch token. According to the priority ranking result, the tasks to be scheduled that meet the release condition are removed from the target queue and controlled to enter the execution state. The tasks to be scheduled that do not meet the release condition are kept in the target queue.

[0075] Specifically, in this embodiment, after generating the corresponding dispatch token, the execution path information in the dispatch token can be parsed first to determine the target queue type corresponding to the task to be scheduled. The execution path information can be used to describe the execution flow of the task to be scheduled under the current hot constraint state. Different execution path information can correspond to different task admission strategies and queue scheduling methods. For example, for a task to be scheduled with a low current hot disturbance level and a high dispatch permission level, an immediate execution path can be used; while for a task to be scheduled with a high current hot load or a target edge computing node in a strong hot coupling state, a delayed execution path can be used. Based on the execution path information, the target queue type corresponding to the task to be scheduled can be determined. The target queue type can include at least an immediate execution queue and a delayed execution queue. The immediate execution queue is used to cache tasks to be scheduled that meet the current execution conditions; the delayed execution queue is used to cache tasks to be scheduled that have not yet met the release conditions.

[0076] After determining the target queue, the tasks to be scheduled can be written into the corresponding target queue according to the admission constraints in the dispatch token. Admission constraints can be used to limit the thermal state conditions that a task must meet before entering the execution state. Different tasks may have different admission constraints. For example, some tasks may require the current thermal state of the target edge computing node to be below a preset thermal threshold before entering the execution state; others may require the thermal coupling effect of adjacent edge computing nodes to decrease to a preset range before execution can proceed.

[0077] Furthermore, within a preset scheduling period, the corresponding dispatch gating function can be called one by one for each task in the target queue to be scheduled. During the call, the admission constraint information corresponding to the task to be scheduled can be used to determine whether the release condition is met. The release condition can be related to the current node's thermal state, the degree of thermal coupling impact, the heat dissipation recovery status, and the current scheduling load status.

[0078] Once a scheduled task is detected to meet its release conditions, the token attributes in the corresponding dispatch token can be read, and the scheduled tasks meeting the release conditions can be prioritized based on these token attributes. This prioritization process determines the execution order of multiple scheduled tasks that meet the release conditions, preventing multiple high-heat-load tasks from entering the execution state simultaneously within the same scheduling cycle. For example, when multiple scheduled tasks simultaneously meet their release conditions, tasks with lower heat response or lower current thermal disturbance risk can be prioritized for execution. After prioritization, scheduled tasks meeting the release conditions can be removed from their corresponding target queues according to the ranking results, and their entry into the execution state can be controlled. Scheduled tasks that do not meet the release conditions remain in the target queue, and release condition checks continue in subsequent scheduling cycles.

[0079] This embodiment allows tasks to be scheduled not to be executed immediately after entering the queue, but rather the timing of task release is dynamically determined based on the current thermal state changes. This reduces the heat accumulation phenomenon caused by the concentrated execution of high-heat-load tasks within a local time window and improves the stability of the task scheduling process under thermal constraints.

[0080] See Figure 4 This application also provides an embodiment of an edge computing node task scheduling device based on thermal coupling constraints, applied in a distributed computing environment containing at least one target edge computing node and at least one adjacent edge computing node, the device comprising: The parameter construction unit 401 is used to construct thermal coupling characterization parameters based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing node; The load characterization value generation unit 402 is used to extract task feature parameters of the task to be scheduled, and generate task thermal load characterization value based on the expected power consumption intensity in the task feature parameters. The thermal shadow segment mapping unit 403 is used to map the task to be scheduled to at least two different thermal shadow segments based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter. Different thermal shadow segments correspond to different thermal disturbance level ranges. The function construction unit 404 is used to construct corresponding dispatch gating functions for the thermal shadow segments respectively. The dispatch gating functions generate dispatch permission based on the task thermal load characterization value and the thermal coupling characterization parameter. The token generation unit 405 is used to generate a corresponding distribution token for each task to be scheduled based on the distribution permission level. The queue control unit 406 is used to map each scheduled task to a target queue of the target edge computing node based on the dispatch token, and to control the scheduled task to enter the execution state from the queue according to the release condition of the dispatch gate function.

[0081] Optionally, the thermal shadow segment mapping unit 403 is specifically used for: Calculate the node thermal coupling influence factor of the target edge computing node in the current scheduling cycle based on the set of thermal coupling characterization parameters. The task thermal load characterization value is coupled with the node thermal coupling influence factor to obtain a comprehensive thermal load index, which is used to characterize the degree of comprehensive thermal disturbance caused by the task when it is executed on the target edge computing node. Based on the heat dissipation margin parameters of the target edge computing node and the preset thermal safety constraints, at least two segmented thresholds are generated for the comprehensive thermal load index. The comprehensive heat load index is compared with the segmented threshold, and the heat disturbance level range to which the task to be scheduled belongs is determined based on the comparison result; Based on the thermal disturbance level range, assign a corresponding thermal shadow segment identifier to the task to be scheduled.

[0082] Optionally, function building unit 404 is specifically used for: Based on the thermal disturbance level range corresponding to the thermal shadow segment, a corresponding initial gating parameter set is determined, wherein the gating parameter set includes at least a basic permissible threshold and an adjustment coefficient; Extract the degree of thermal influence between the target edge computing node and its adjacent edge computing nodes, and generate a thermal coupling correction factor; The task thermal load characterization value is input into a preset load response mapping relationship to obtain the thermal response amount generated by the task to the target edge computing node; The initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity are coupled to obtain the dispatch gating function; The task thermal load characterization value is substituted into the dispatch gating function for evaluation to obtain the corresponding dispatch permission degree.

[0083] Function building unit 404 is specifically used for: Based on the initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity, corresponding gating influence components are constructed respectively. Nonlinear modulation is applied to the gated influence component, so that the dispatch gate function exhibits nonlinear suppression characteristics when thermal coupling is enhanced or the task thermal load increases; The gated influence components after nonlinear modulation are combined and operated to generate the dispatch gated function.

[0084] Optionally, the token generation unit 405 is specifically used for: Determine the permission level to which the task to be scheduled belongs; Based on the permission level, a corresponding token attribute is determined for the task to be scheduled; The token attributes are modified based on the current thermal environment; The modified token attributes are encoded to generate a dispatch token associated with the task to be scheduled.

[0085] Optionally, the queue control unit 406 is specifically used for: The execution path information in the dispatch token is parsed, and the target queue type corresponding to the task to be scheduled is determined based on the execution path information. The target queue type includes at least an immediate execution queue and a delayed execution queue. Write the task to be scheduled into the target queue according to the admission constraint information in the dispatch token; Within the preset scheduling period, the corresponding dispatch gating function is called one by one for the scheduled tasks in the target queue, and the release condition is determined based on the admission constraint information. For scheduled tasks that meet the release conditions, priority is sorted according to the token attributes in the dispatch token; Based on the priority sorting result, the scheduled tasks that meet the release conditions are removed from the target queue and the scheduled tasks are controlled to enter the execution state; Tasks that have not met the release conditions are kept in the target queue.

[0086] Optionally, the thermal shadow segment is a thermal disturbance mapping interval divided based on the task thermal load characterization value and the degree of thermal influence between nodes.

[0087] Optionally, the task thermal load characterization value is obtained based on the mapping relationship between the task's computational load estimate and power consumption intensity.

[0088] Please see Figure 5 This application also provides an edge computing node task scheduling device based on thermal coupling constraints, comprising: Processor 501, memory 502, input / output unit 503, bus 504; The processor 501 is connected to the memory 502, the input / output unit 503, and the bus 504; The memory 502 stores a program, and the processor 501 calls the program to execute any of the methods described above.

[0089] This application also relates to a computer-readable storage medium on which a program is stored, which, when run on a computer, causes the computer to perform any of the methods described above.

[0090] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0091] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0093] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0094] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A task scheduling method for edge computing nodes based on thermal coupling constraints, characterized in that, The method, applied in a distributed computing environment comprising at least one target edge computing node and at least one adjacent edge computing node, includes: Based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing nodes, thermal coupling characterization parameters are constructed; Extract task feature parameters from the task to be scheduled, and generate task thermal load characterization values ​​based on the expected power consumption intensity in the task feature parameters; Based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter, the task to be scheduled is mapped to at least two different thermal shadow segments, and different thermal shadow segments correspond to different thermal disturbance level ranges. For each of the thermal shadow segments, a corresponding dispatch gating function is constructed. The dispatch gating function generates a dispatch permission degree based on the task thermal load characterization value and the thermal coupling characterization parameter. Based on the dispatch permission level, generate a corresponding dispatch token for each task to be scheduled. Based on the dispatch token, each task to be scheduled is mapped to the target queue of the target edge computing node, and the task to be scheduled is controlled to enter the execution state from the queue according to the release condition of the dispatch gate function.

2. The edge computing node task scheduling method based on thermal coupling constraints as described in claim 1, characterized in that, Based on the task thermal load characterization value and the inter-node thermal influence degree characterized by the thermal coupling characterization parameter, the task to be scheduled is mapped to at least two different thermal shadow segments, each corresponding to a different thermal disturbance level range, including: Calculate the node thermal coupling influence factor of the target edge computing node in the current scheduling cycle based on the set of thermal coupling characterization parameters. The task thermal load characterization value is coupled with the node thermal coupling influence factor to obtain a comprehensive thermal load index, which is used to characterize the degree of comprehensive thermal disturbance caused by the task when it is executed on the target edge computing node. Based on the heat dissipation margin parameters of the target edge computing node and the preset thermal safety constraints, at least two segmented thresholds are generated for the comprehensive thermal load index. The comprehensive heat load index is compared with the segmented threshold, and the heat disturbance level range to which the task to be scheduled belongs is determined based on the comparison result; Based on the thermal disturbance level range, assign a corresponding thermal shadow segment identifier to the task to be scheduled.

3. The edge computing node task scheduling method based on thermal coupling constraints as described in claim 1, characterized in that, For each of the thermal shadow segments, a corresponding dispatch gating function is constructed. The dispatch gating function generates a dispatch permission level based on the task thermal load characterization value and the thermal coupling characterization parameter, including: Based on the thermal disturbance level range corresponding to the thermal shadow segment, a corresponding initial gating parameter set is determined, wherein the gating parameter set includes at least a basic permissible threshold and an adjustment coefficient; Extract the degree of thermal influence between the target edge computing node and its adjacent edge computing nodes, and generate a thermal coupling correction factor; The task thermal load characterization value is input into a preset load response mapping relationship to obtain the thermal response amount generated by the task to the target edge computing node; The initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity are coupled to obtain the dispatch gating function; The task thermal load characterization value is substituted into the dispatch gating function for evaluation to obtain the corresponding dispatch permission degree.

4. The edge computing node task scheduling method based on thermal coupling constraints as described in claim 1, characterized in that, The initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity are coupled to obtain the dispatch gating function, including: Based on the initial gating parameter set, the thermal coupling correction factor, and the thermal response quantity, corresponding gating influence components are constructed respectively. Nonlinear modulation is applied to the gated influence component, so that the dispatch gate function exhibits nonlinear suppression characteristics when thermal coupling is enhanced or the task thermal load increases; The gated influence components after nonlinear modulation are combined and operated to generate the dispatch gated function.

5. The edge computing node task scheduling method based on thermal coupling constraints as described in claim 1, characterized in that, The step of generating corresponding dispatch tokens for each task to be scheduled based on the dispatch permission level includes: Determine the permission level to which the task to be scheduled belongs; Based on the permission level, a corresponding token attribute is determined for the task to be scheduled; The token attributes are modified based on the current thermal environment; The modified token attributes are encoded to generate a dispatch token associated with the task to be scheduled.

6. The edge computing node task scheduling method based on thermal coupling constraints as described in claim 1, characterized in that, The step of mapping each task to be scheduled to a target queue on the target edge computing node based on the dispatch token, and controlling the task to be scheduled to enter the execution state from the queue according to the release condition of the dispatch gating function, includes: The execution path information in the dispatch token is parsed, and the target queue type corresponding to the task to be scheduled is determined based on the execution path information. The target queue type includes at least an immediate execution queue and a delayed execution queue. Write the task to be scheduled into the target queue according to the admission constraint information in the dispatch token; Within the preset scheduling period, the corresponding dispatch gating function is called one by one for the scheduled tasks in the target queue, and the release condition is determined based on the admission constraint information. For scheduled tasks that meet the release conditions, priority is sorted according to the token attributes in the dispatch token; Based on the priority sorting result, the scheduled tasks that meet the release conditions are removed from the target queue and the scheduled tasks are controlled to enter the execution state; Tasks that have not met the release conditions are kept in the target queue.

7. The edge computing node task scheduling method based on thermal coupling constraints according to any one of claims 1 to 6, characterized in that, The thermal shadow segment is a thermal disturbance mapping interval divided based on the task thermal load characterization value and the degree of thermal influence between nodes.

8. The edge computing node task scheduling method based on thermal coupling constraints according to any one of claims 1 to 6, characterized in that, The task thermal load characterization value is obtained based on the mapping relationship between the task's computational load estimate and power consumption intensity.

9. An edge computing node task scheduling device based on thermal coupling constraints, characterized in that, The apparatus, applied in a distributed computing environment comprising at least one target edge computing node and at least one adjacent edge computing node, includes: The parameter construction unit is used to construct thermally coupled characterization parameters based on the spatial adjacency relationship between the target edge computing node and the adjacent edge computing node; The load characterization value generation unit is used to extract task feature parameters from the task to be scheduled, and generate task thermal load characterization values ​​based on the expected power consumption intensity in the task feature parameters. The thermal shadow segment mapping unit is used to map the task to be scheduled to at least two different thermal shadow segments based on the task thermal load characterization value and the degree of inter-node thermal influence characterized by the thermal coupling characterization parameter. Different thermal shadow segments correspond to different thermal disturbance level ranges. The function construction unit is used to construct corresponding dispatch gating functions for the thermal shadow segments respectively. The dispatch gating functions generate dispatch permission based on the task thermal load characterization value and the thermal coupling characterization parameter. The token generation unit is used to generate a corresponding distribution token for each task to be scheduled based on the distribution permission level. The queue control unit is used to map each scheduled task to a target queue of the target edge computing node based on the dispatch token, and to control the scheduled task to enter the execution state from the queue according to the release condition of the dispatch gate function.

10. An edge computing node task scheduling device based on thermal coupling constraints, characterized in that, The device includes: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a program that, when executed on a computer, performs the method as described in any one of claims 1 to 8.

Images