Data processing method and device and electronic equipment

Abstract

The invention provides a data processing method and device and electronic equipment, and the method comprises the steps: determining whether an adjustment triggering condition is satisfied or not based on a first performance change rate and a first temperature change rate of the electronic equipment; determining a task control instruction based on a first performance index of the electronic equipment in response to the condition that the adjustment triggering condition is met, and determining a temperature regulation instruction based on a predicted performance index of the electronic equipment; task execution of the electronic equipment is scheduled based on the task control instruction, and a refrigeration system is controlled based on a temperature regulation and control instruction; wherein the time scale of the task control instruction is smaller than the time scale of the temperature regulation and control instruction; the time scale comprises a time unit of a duration required for executing the control instruction or a time unit of a duration required for responding to the control instruction.

Description

Technical Field

[0001] This disclosure relates to the field of electronic equipment technology, and in particular to a data processing method, apparatus and electronic equipment. Background Technology

[0002] With the rapid development of artificial intelligence and high-performance computing applications, data centers are facing increasingly severe challenges in energy consumption and thermal management. The task scheduling of information technology (IT) workloads fluctuates rapidly on a second or minute-by-minute basis, while the response of cooling systems, limited by thermal inertia, often takes several minutes or even longer. This inherent contradiction of "fast IT, slow cooling" stems from the lack of intelligent control mechanisms in existing control systems capable of simultaneously coordinating two time scales. Due to this time scale mismatch and the absence of dual-time-scale controllers, the system suffers from response lag, command conflicts, low energy efficiency (resulting in energy savings on one side while energy consumption on the other), overcooling, localized hotspots, and resource waste. Summary of the Invention

[0003] This disclosure provides a data processing method, apparatus, and electronic device to at least solve the above-mentioned technical problems existing in the prior art.

[0004] According to a first aspect of this disclosure, a data processing method is provided, comprising: Based on the first performance change rate and the first temperature change rate of the electronic device, determine whether the adjustment trigger condition is met; In response to the fulfillment of the adjustment triggering condition, a task control command is determined based on a first performance index of the electronic device, and a temperature control command is determined based on a predicted performance index of the electronic device. The task execution of electronic devices is scheduled based on the task control instructions, and the refrigeration system is controlled based on the temperature regulation instructions. The time scale of the task control command is smaller than that of the temperature control command; the time scale includes the time unit required to execute the control command, or the time unit required to respond to the control command.

[0005] According to a second aspect of this disclosure, a data processing apparatus is provided, the apparatus comprising: The judgment unit is used to determine whether the adjustment triggering condition is met based on the first performance change rate and the first temperature change rate of the electronic device. The determining unit is configured to, in response to the satisfaction of the adjustment triggering condition, determine a task control command based on a first performance index of the electronic device, and simultaneously determine a temperature control command based on a predicted performance index of the electronic device. An execution unit is used to schedule the execution of tasks of electronic devices based on the task control instructions and to control the refrigeration system based on temperature control instructions. The time scale of the task control command is smaller than that of the temperature control command; the time scale includes the time unit required to execute the control command, or the time unit required to respond to the control command.

[0006] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the methods of this disclosure.

[0007] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0008] The above and other objects, features, and advantages of this disclosure will become readily apparent from the following detailed description of exemplary embodiments, taken in conjunction with the accompanying drawings. Several embodiments of this disclosure are illustrated in the drawings by way of example and not limitation, in which: In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0009] Figure 1 A schematic diagram of a first optional flow of the data processing method provided in this embodiment of the present disclosure is shown; Figure 2 A schematic diagram of a second optional flow of the data processing method provided in an embodiment of this disclosure is shown; Figure 3 A schematic diagram of a third optional flow of the data processing method provided in this embodiment of the present disclosure is shown; Figure 4 A schematic diagram of a fourth optional flow of the data processing method provided in this disclosure embodiment is shown; Figure 5 A schematic diagram of a fifth optional process for the data processing method provided in an embodiment of this disclosure is shown; Figure 6 The illustration shows an application scenario provided by an embodiment of this disclosure; Figure 7 The diagram shown is a scheduling schematic of an embodiment of this disclosure; Figure 8 A schematic diagram of an optional structure of the data processing apparatus provided in an embodiment of this disclosure is shown; Figure 9 A schematic diagram of the composition structure of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0010] To make the objectives, features, and advantages of this disclosure more apparent and understandable, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0011] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0012] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this disclosure described herein can be implemented in an order other than that illustrated or described herein.

[0013] Unless otherwise defined, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used in this disclosure is for the purpose of describing embodiments of this disclosure only and is not intended to be limiting of this disclosure.

[0014] It should be understood that in the various embodiments of this disclosure, the sequence number of each implementation process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0015] Before providing a further detailed description of the embodiments of this disclosure, the nouns and terms involved in the embodiments of this disclosure will be explained, and the nouns and terms involved in the embodiments of this disclosure shall be interpreted as follows.

[0016] Heterogeneous systems refer to integrated systems consisting of discrete event-driven computing systems (such as job schedulers) and continuous-time dynamic physical cooling systems (such as liquid-cooled CDUs).

[0017] Dual time scales refer to a first decision time scale (ΔT_IT, milliseconds, seconds) that matches the frequency of changes in the computational load, and a second decision time scale (ΔT_C, tens of seconds, minutes, hours) that matches the thermal inertia of the cooling system, where ΔT_C > ΔT_IT.

[0018] Due to the lack of dynamic collaborative judgment and decision-making capabilities across two time scales, existing systems in related technologies typically suffer from the following problems: Frequent invalid commands and low control efficiency: In fixed-cycle control mode, the system cannot distinguish between fast-changing events and slow-adjustment processes. When a cooling command is issued, due to the slow thermal response, the controller continues to issue new commands based on old data before the state is updated, resulting in a series of invalid or even contradictory operations. This reflects the system's lack of a decision-making rhythm mechanism that can be triggered on demand and dynamically adapted, failing to match the control rhythm to the actual response speed.

[0019] The system exhibits continuous oscillations and poor stability: lacking predictive control with bidirectional coordination, the system can only make "follow-up" adjustments based on minor deviations in the current state. For example, it raises the setting when the water temperature drops slightly and lowers it hastily when the temperature rises again, causing the cooling parameters to oscillate repeatedly around the set value. This instability is essentially due to the lack of a cross-system forward-looking coordination and smooth transition mechanism, making it impossible to synchronize and buffer commands.

[0020] Continuous energy waste: To compensate for the lag in response, the system can only resort to "overcooling," providing cooling capacity that exceeds actual demand for an extended period.

[0021] Thermal safety runaway risk: When IT load suddenly increases, the cooling system may not respond in time and may not be able to adjust the computing tasks in time, leading to local overheating.

[0022] In related technologies, control solutions for thermal management in data centers have the following limitations: Fixed-frequency control mechanisms are rigid: Most existing systems use a fixed time interval control cycle, adjusting at the same pace regardless of drastic fluctuations in IT load or stable system operation. This "one-size-fits-all" decision-making model cannot adapt to dynamically changing system characteristics: shortening the cycle to pursue rapid response easily leads to oscillations, while lengthening the cycle to maintain stability results in response lag. The fundamental reason lies in the lack of a dual-timescale mechanism that can adaptively adjust the decision-making rhythm according to the real-time state of the system.

[0023] Rule-based suppression strategies are passive and rigid: To alleviate the oscillation problem caused by frequent adjustments, a common practice is to introduce rules such as dead zones or rate limits for manual intervention. This method is essentially a static, passive instruction filter. While it can reduce unnecessary actions, it cannot intelligently determine when to intervene decisively and when to remain silent. It solves the problem of "too many actions," but introduces the new risk of "not acting when it should," reflecting its lack of ability to dynamically adjust decision sensitivity based on the importance of events.

[0024] Limitations of Single-Time-Scale Optimization: Whether it's traditional PID control or advanced methods that partially incorporate artificial intelligence, their decision-making processes are mostly still limited to a single time scale. They either treat IT scheduling and cooling regulation separately, or they respond to fast and slow dynamics in a unified manner at the same frequency, failing to establish a collaborative framework across time scales. This inherent flaw makes it difficult to fundamentally bridge the gap between IT's second-level fluctuations and cooling's minute-level response.

[0025] In summary, related technologies either completely ignore the core contradiction of time scale mismatch or only adopt local, one-way repair strategies. Due to the lack of an adaptive dual-time-scale control mechanism capable of unified perception, dynamic decision-making, and bidirectional coordination, the system can never truly achieve global optimization of safety and energy efficiency. Therefore, there is an urgent need for an intelligent collaborative control method that can integrate IT and cooling systems while considering both response speed and operational stability.

[0026] In view of the deficiencies existing in the related technologies, this disclosure provides a data processing method to at least solve some or all of the above-mentioned technical problems.

[0027] Figure 1 A schematic diagram of a first alternative flow of the data processing method provided in this disclosure embodiment is shown, and the steps will be described accordingly.

[0028] Step S101: Based on the first performance change rate and the first temperature change rate of the electronic device, determine whether the adjustment trigger condition is met.

[0029] In some embodiments, the electronic device may be an electronic device corresponding to a data center (such as a computing device, storage device, network device, power device, and management and monitoring device), or it may be at least one of a server, an AI all-liquid silent GPU workstation, a personal computer host, a high-bandwidth memory computing device, a high-performance computing device, and an edge computing node device.

[0030] In some embodiments, the first performance change rate of the electronic device may include at least one of the following: power change rate over a first time interval, IT power supply change rate (dynamic response speed of output capability), server utilization, task queue status (including the number of tasks to be executed in the queue, priority distribution, processing progress, etc.), and system input status (real-time changes in external input conditions that drive the operation of the electronic device).

[0031] In some embodiments, the first temperature change rate of the electronic device may include the temperature change rate of each key heat-generating component (such as CPU, GPU, video memory, and chipset) in the electronic device within a first time interval, the water flow rate of the coolant in the cooling system, and the water temperature change rate of the coolant in the cooling system.

[0032] In some embodiments, the carrier implementing the data processing method (hereinafter referred to as the carrier) may determine whether the adjustment trigger condition is met based on the magnitude between the first performance change rate and the performance change threshold; or, based on the magnitude between the first temperature change rate and the temperature change threshold; optionally, it may also determine whether the adjustment trigger condition is met based on the magnitude between the first performance change rate and the first temperature change rate and the comprehensive change threshold.

[0033] Specifically, the carrier can determine whether the adjustment triggering condition is met based on the power change rate in the first performance change rate; for example, if the power change rate is greater than the power change threshold, then the adjustment triggering condition is met; wherein, the performance change threshold includes the power change threshold.

[0034] Alternatively, specifically, the carrier can determine whether the adjustment triggering condition is met based on the temperature change rate of the key heating component in the first temperature change rate; for example, if the temperature change rate is greater than the temperature change threshold, then the adjustment triggering condition is determined to be met.

[0035] Alternatively, the carrier can quantify each parameter in the first performance change rate, quantizing the value of each parameter to a score of a uniform dimension, and then determine the score of the first performance change rate based on the weight of each parameter; if the score of the first performance change rate is greater than the performance change threshold, then the adjustment trigger condition is determined to be met.

[0036] Alternatively, the carrier can quantify each parameter in the first temperature change rate, quantizing the value of each parameter to a score of a uniform dimension, and then determine the score of the first temperature change rate based on the weight of each parameter; if the score of the first performance change rate is greater than the temperature change threshold, then the adjustment trigger condition is determined to be met.

[0037] Alternatively, the carrier can quantify each parameter in the first performance change rate and the first temperature change rate to obtain a comprehensive score of the first performance change rate and the first temperature change rate. If the comprehensive score is greater than the comprehensive change threshold, then the adjustment trigger condition is determined to be met.

[0038] The adjustment triggering conditions include triggering conditions for adjustment commands for discrete tasks and continuous temperature adjustment commands; that is, if the adjustment triggering conditions are met, adjustments can be made to both discrete tasks and continuous temperature (or temperature control system). The adjustment includes issuing adjustment commands.

[0039] The carrier can be a computer program, electronic circuit, database, mobile application, electronic device, cloud computing platform, distributed system, artificial intelligence framework, mathematical model, automation tool and microcontroller, etc., which are software or hardware capable of implementing algorithms and methods.

[0040] In step S102, in response to the satisfaction of the adjustment triggering condition, a task control command is determined based on the first performance index of the electronic device, and a temperature control command is determined based on the predicted performance index of the electronic device.

[0041] In some embodiments, in response to the fulfillment of the adjustment triggering condition, the carrier determines a task control instruction based on a first performance indicator of the electronic device. The first performance indicator of the electronic device may include at least one of the following within a first time interval: processor utilization (CPU utilization), task execution pressure, task execution status, and load of any node. The task control instruction may include suspending the task, scheduling the task to another node (such as a node with sufficient cooling capacity), or adjusting the task priority to advance or postpone the task's execution.

[0042] In other embodiments, in response to the fulfillment of the adjustment triggering condition, the carrier determines a temperature regulation command simultaneously with the task control command. Specifically, the carrier can determine, based on the predicted performance indicators of the electronic device, the temperature changes of key heat-generating components of the electronic device, the temperature and flow rate of the coolant in the heat dissipation system, within a second time interval without any adjustment or intervention to the heat dissipation system, and determine the temperature regulation command based on the temperature changes of key heat-generating components of the electronic device, the temperature and flow rate of the coolant in the heat dissipation system, within the second time interval. The temperature regulation command may include increasing the flow rate, decreasing the flow rate, increasing the flow rate, decreasing the flow rate, decreasing the inlet water temperature, increasing the inlet water temperature, accelerating the rate of decrease of the inlet water temperature, decreasing the rate of decrease of the inlet water temperature, increasing the cold plate bonding pressure, decreasing the cold plate bonding pressure, increasing the coolant level, decreasing the coolant level, increasing the fan speed, decreasing the fan speed, increasing the cooling tower fan frequency, and decreasing the cooling tower fan frequency, at least one of these. The second time interval is after the first time interval and may be a time interval that has not yet occurred.

[0043] The time scale of the task control instruction is smaller than that of the temperature control instruction; the time scale includes the time unit required to execute the control instruction, or the time unit required to respond to the control instruction. For example, the time scale of the task control instruction is on the order of milliseconds or seconds, while the time dimension of the temperature control instruction is on the order of tens of seconds, minutes, or even hours.

[0044] Step S103: Schedule the task execution of the electronic device based on the task control command, and control the refrigeration system based on the temperature regulation command.

[0045] In some embodiments, the carrier executes each instruction at the time corresponding to each instruction, that is, it schedules the task execution of the electronic device based on the task control instruction and controls the cooling system based on the temperature regulation instruction.

[0046] Thus, the data processing method provided in this disclosure bridges the time-scale gap between rapid IT fluctuations and slow cooling responses, dynamically coordinating discrete IT scheduling decisions with continuous cooling system adjustments. This upgrades the control mode from a traditional "single-scale passive response" to "two-way collaborative active regulation," thereby achieving optimal global energy efficiency. Specifically, it dynamically adjusts the decision-making rhythm of both the fast (IT decision-making) and slow (cooling regulation) time scales based on the system's actual load fluctuations and thermal state. It accelerates response during high-dynamic phases and extends silence during stable phases, precisely matching the system's inherent dynamic characteristics. This fundamentally avoids invalid operations and command conflicts, significantly improving control accuracy, efficiency, and overall robustness.

[0047] Figure 2A second alternative flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0048] Step S201: Based on the first performance change rate and the first temperature change rate of the electronic device, determine whether the adjustment trigger condition is met.

[0049] In some embodiments, in response to the first performance change rate being greater than a performance change threshold, or the first temperature change rate being greater than a temperature change threshold, it is confirmed that the event corresponding to the current moment meets the adjustment trigger condition. The specific determination method is the same as step S101, and will not be repeated here.

[0050] In other embodiments, if the first power change rate, which includes the first performance change rate, is greater than a power change threshold, or the first temperature change rate is greater than a temperature change threshold, then the carrier determines that an event exists. The priority of the event is further determined, and based on the event's priority and a priority threshold, it is determined whether the event meets the adjustment triggering condition. Specifically, the carrier can identify the event source based on the first performance change rate and the first temperature change rate. The event priority is determined based on the event source, performance change rate or temperature change rate, event duration, and the weights of each parameter; in response to the event's priority being greater than the priority threshold, it is determined that the event corresponding to the current moment meets the adjustment triggering condition.

[0051] Specifically, the carrier can compare a first performance change rate with the performance change rate when no event occurs. If the two are similar (the absolute value of the difference is less than a fifth threshold), the event source is determined to be non-task-side; if the difference is large (the absolute value of the difference is greater than the fifth threshold), the event source is determined to be task-side. Alternatively, the carrier can compare a first temperature change rate with the temperature change rate when no event occurs. If the two are similar (the absolute value of the difference is less than a sixth threshold), the event source is determined to be non-heat dissipation-side; if the difference is large (the absolute value of the difference is greater than the sixth threshold), the event source is determined to be heat dissipation-side. The event source includes reasons that cause the first performance change rate to exceed the performance change threshold, or reasons that cause the first temperature change rate to exceed the temperature change threshold. It can be task-side (IT side) or heat dissipation-side (cooling side). The fifth and sixth thresholds can be set according to actual needs.

[0052] In some embodiments, the carrier determines the priority of an event based on the event source, performance change rate or temperature change rate, event duration, and the weights of each parameter, including: determining the priority of an event based on the weight corresponding to the event source, the event type coefficient, the amplitude measurement function, the amplitude weight, the event duration, and the duration weight.

[0053] Different event sources have different weights; the event type coefficient is determined based on the event source and is a configurable value; the amplitude measurement function is related to the rate of change; if the event source is the task side, the amplitude measurement function is the function corresponding to the power change rate; if the event source is the heat dissipation side, the amplitude measurement function is the function corresponding to the temperature change rate; the amplitude weight includes the weight of controlling the event change amplitude in the overall score (or priority); the event duration includes the duration of events that cause performance changes or temperature changes; the duration weight includes the weight of controlling the event duration in the overall score (or priority).

[0054] In some embodiments, event priority is used to assess events in real time and determine their importance and urgency.

[0055] In step S202, in response to the satisfaction of the adjustment triggering condition, a task control command is determined based on the first performance index of the electronic device, and a temperature control command is determined based on the predicted performance index of the electronic device.

[0056] The specific steps of step S202 are the same as those of step S102, and will not be repeated here.

[0057] Step S203: Schedule the task execution of the electronic device based on the task control command, and control the refrigeration system based on the temperature regulation command.

[0058] The specific steps of step S203 are the same as those of step S103, and will not be repeated here.

[0059] Thus, the data processing method provided in this disclosure bridges the time-scale gap between rapid IT fluctuations and slow cooling response, dynamically coordinating discrete IT scheduling decisions with continuous cooling system adjustments. This upgrades the control mode from the traditional "single-scale passive response" to "two-way collaborative active regulation," thereby achieving optimal global energy efficiency. Specifically, it monitors the IT power change rate and cooling return water temperature gradient in real time. When a change exceeds a preset threshold, a corresponding event is triggered. Subsequently, a weighted scoring mechanism comprehensively evaluates the importance and urgency of the event, considering factors such as event type, change magnitude, and duration. For highly important events, collaborative decision-making is immediately triggered, significantly improving the accuracy, efficiency, and overall robustness of control.

[0060] Figure 3 A third alternative flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0061] Step S301: Based on the first performance change rate and the first temperature change rate of the electronic device, determine whether the adjustment trigger condition is met.

[0062] The specific steps of step S301 are the same as those of step S101 or step S201, and will not be repeated here.

[0063] Step S302: Determine the task control instructions based on the first performance index of the electronic device.

[0064] In some embodiments, different first performance indicators correspond to different task control instructions.

[0065] Specifically, in response to the power included in the first performance indicator being greater than a first threshold, and the priority of the event included in the first performance indicator being greater than a second threshold, the task corresponding to the event is stopped from being executed and the task corresponding to the event is suspended.

[0066] Alternatively, specifically, in response to the first performance metric (including the first node load) being less than a third threshold and the first node resource utilization being less than a fourth threshold, the task corresponding to the event is scheduled to the first node, and the first node executes the task. That is, the node with sufficient cooling capacity and lower current pressure handles the task.

[0067] Alternatively, for scenarios other than the two mentioned above, the carrier can adjust the execution order of tasks, allowing tasks to be processed earlier, later, or maintaining the current execution order.

[0068] Step S303: Determine the temperature control command based on the predicted performance indicators of the electronic device.

[0069] In some embodiments, the carrier determines the predicted power change from the current moment to a first moment based on historical events, historical tasks, and historical performance indicators of the electronic device; the first moment is a future moment, i.e. a moment that has not yet occurred.

[0070] Specifically, the carrier can train the model based on historical events, historical tasks, historical performance indicators, and power changes of electronic devices collected before the current moment, so that the model has the ability to predict power changes based on historical events, historical tasks, and historical performance indicators of electronic devices.

[0071] In some embodiments, the carrier determines a cooling parameter adjustment sequence from the current moment to a first moment based on predicted power change, maximum power value, thermal safety scheduling adjustment factor, control gain coefficient, and current temperature control command; the cooling parameter adjustment sequence from the current moment to the first moment is determined as the temperature control command. The maximum power value can be determined based on predicted power change or historical performance indicators; the thermal safety scheduling adjustment factor is used to dynamically adjust the control intensity according to the proximity of the current temperature to a safety threshold; the control gain coefficient is used to control the adjustment step size; the cooling parameter adjustment sequence is a forward-looking control sequence of the heat dissipation system, including adjustment strategies from the current moment to the first moment, such as return water temperature and flow rate setpoints.

[0072] Step S304: Schedule the task execution of the electronic device based on the task control command, and control the refrigeration system based on the temperature regulation command.

[0073] The specific steps of step S304 are the same as those of step S103, and will not be repeated here.

[0074] Thus, the data processing method provided in this disclosure bridges the time-scale gap between rapid IT fluctuations and slow cooling response, dynamically coordinating discrete IT scheduling decisions with continuous cooling system adjustments. This upgrades the control mode from the traditional "single-scale passive response" to "two-way collaborative active regulation," thereby achieving optimal global energy efficiency. Specifically, it monitors the IT power change rate and cooling return water temperature gradient in real time. When a change exceeds a preset threshold, a corresponding event is triggered. Subsequently, a weighted scoring mechanism comprehensively evaluates the importance and urgency of the event, considering factors such as event type, change magnitude, and duration. For highly important events, collaborative decision-making is immediately triggered, significantly improving the accuracy, efficiency, and overall robustness of control.

[0075] Figure 4 A fourth optional flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0076] Step S401: Determine whether the adjustment trigger condition is met.

[0077] In some embodiments, the carrier can determine task control commands and temperature regulation commands based on adjusting the trigger cycle.

[0078] Specifically, in response to the arrival of the adjustment trigger cycle, it is determined that the adjustment trigger condition is met. The carrier determines the task control command based on the first performance index of the electronic device and the temperature control command based on the predicted performance index of the electronic device. The adjustment trigger cycle is determined based on the minimum decision interval, the stability influence coefficient, the standard deviation of the task-side power, the standard deviation of the heat dissipation-side power, the maximum value of the task-side power, and the maximum value of the heat dissipation-side power.

[0079] In other embodiments, the carrier can also monitor the performance change rate and temperature change rate of the electronic device, and determine whether the adjustment trigger condition is met based on the performance change rate and performance change threshold, the temperature change rate and temperature change threshold. The specific determination method is as described in steps S101, S201 or S301, and will not be repeated here.

[0080] In other words, the carrier can adjust the triggering based on the adjustment triggering period (i.e., determine and execute task control instructions and temperature control instructions); or it can adjust the triggering based on the performance change rate and temperature change rate if the adjustment triggering period has not been reached.

[0081] In step S402, in response to the satisfaction of the adjustment triggering condition, a task control command is determined based on the first performance index of the electronic device, and a temperature control command is determined based on the predicted performance index of the electronic device.

[0082] The specific steps of step S402 are the same as those of steps S102, S302 to S303, and will not be repeated here.

[0083] Step S403: Schedule the task execution of the electronic device based on the task control command, and control the refrigeration system based on the temperature regulation command.

[0084] The specific steps of step S403 are the same as those of step S103, and will not be repeated here.

[0085] In some embodiments, after performing one of steps S103, S203, S304 and S403, the method may further include steps S404 to S406.

[0086] Step S404: Update the priority threshold.

[0087] In some embodiments, if the adjustment trigger cycle has not been reached, the carrier can monitor the performance indicators and temperature information of the electronic device, and determine whether to trigger an event based on power change thresholds and temperature change thresholds; if triggered, the priority of the event is determined, and whether the adjustment trigger condition is met is determined based on the event priority and priority threshold. The priority of the event can be determined based on the event source, performance change rate or temperature change rate, event duration, and the weight of each parameter.

[0088] In some embodiments, after scheduling the task execution of the electronic device based on the task control command and controlling the cooling system based on the temperature regulation command, the carrier continuously collects the performance indicators and temperature information of the electronic device, and updates the weights of each parameter required to determine the priority of the event, as well as the priority threshold, based on the performance indicators and temperature information.

[0089] In some embodiments, the carrier can update the weight of each parameter based on the comprehensive energy consumption cost at the second time point, the comprehensive energy consumption cost at the third time point before the second time point, the control adjustment step size, and a preset adjustment amount. The second and third times are times after the first time point, and can be any time after executing step S403. The updated weight of each parameter is used to determine the priority of the event corresponding to the fourth time point. The fourth time point is a time after the second time point.

[0090] The parameters may include the weight corresponding to the event source, the magnitude weight, and the duration weight (i.e., the weights involved in determining the priority of the event in step S201).

[0091] In some embodiments, the comprehensive energy consumption cost at the second moment is determined based on the task-side power, heat dissipation-side power, a first coefficient, and a second coefficient weighted at the second moment. The sum of the first coefficient and the second coefficient is 1. The comprehensive energy consumption cost at the second moment and the comprehensive energy consumption cost at the third moment prior to the second moment are used to determine the direction of performance improvement. If the comprehensive energy consumption cost at the previous moment is greater than the comprehensive energy consumption cost at the next moment, it indicates that the energy consumption cost has decreased, and optimization can continue along the current adjustment direction; conversely, if the comprehensive energy consumption cost at the previous moment is less than the comprehensive energy consumption cost at the next moment, it indicates that the energy consumption cost has increased, and optimization can continue along the opposite direction of the current adjustment direction. The preset adjustment amount is a preset fixed adjustment amount, which can ensure the stability and controllability of the weight optimization process.

[0092] In some embodiments, the carrier updates the priority threshold based on the priority threshold corresponding to the second time moment, the number of events that meet the adjustment triggering conditions within the third time interval, the expected maximum event frequency reference value within the third time interval, and a dynamic adjustment coefficient. The updated priority threshold is used to determine whether the event at the fourth time moment meets the adjustment triggering conditions. The third time interval is the time interval before the second time moment.

[0093] The dynamic adjustment coefficient is used to automatically increase its value to filter out interference and enhance stability when events are too frequent; conversely, it decreases its value to improve the system's response sensitivity when events are less frequent.

[0094] Step S405: Update the control gain coefficient.

[0095] In some embodiments, the control gain coefficient is a parameter used in step S303 when determining the temperature control command, and is used to control the adjustment step size.

[0096] In some embodiments, the carrier corrects the control gain coefficient by monitoring the coordinated power changes on the task side and the heat dissipation side in order to suppress oscillations and optimize response speed.

[0097] Specifically, the carrier determines the control gain coefficient at the fourth time based on the control gain coefficient at the second time moment, the power change on the mission side during the third time interval, the power change on the heat dissipation side during the third time interval, the synergy effect indicator, and the gain adjustment coefficient.

[0098] The synergy effect indicator is determined based on the power change on the task side and the power change on the heat dissipation side within the third time interval. When the two changes in opposite directions, it indicates that the control may have a counterproductive effect, and the gain control gain coefficient needs to be increased; when they are in the same direction, it may be overshoot, and the control gain coefficient needs to be decreased.

[0099] Step S406: Update and adjust the trigger cycle.

[0100] In some embodiments, the carrier determines the adjustment triggering period based on the minimum decision interval, stability influence coefficient, standard deviation of task-side power, standard deviation of heat dissipation-side power, maximum value of task-side power, and maximum value of heat dissipation-side power.

[0101] In some embodiments, as the electronic device operates, the execution status of various tasks, and the electronic device executes task control commands and temperature regulation commands, the power change rate and temperature change rate change accordingly. The carrier can update the adjustment trigger cycle based on the duration of stable operation of the electronic device, the silent adjustment coefficient, and the maximum stable duration, so that the adjustment trigger cycle is longer when the system is stable, reducing load and energy consumption; and the adjustment cycle can be shortened when the system is unstable, improving overall robustness.

[0102] Thus, through the data processing method provided in this embodiment, by real-time monitoring of key indicators such as power changes and cooling temperature gradients on the task side, high-priority events (such as sudden increases in tasks) are automatically identified, and collaborative decision-making is immediately triggered. During stable system operation, the decision interval is automatically extended, entering a low-overhead "energy-saving silent mode." This mechanism enables the control system to respond quickly at critical moments and reduce intervention during idle times, achieving a balance between responsiveness and control efficiency. At each collaborative decision-making moment, bidirectional control commands are generated synchronously: precise job scheduling intervention is implemented on the task side, such as delaying submission or guiding tasks to nodes with sufficient cooling capacity; a set of forward-looking parameter adjustment sequences is output for the refrigeration system, such as gradually adjusting the return water temperature and flow rate to respond in advance to predicted load changes. This mechanism achieves a bidirectional closed loop at the action level, effectively suppressing system oscillations and avoiding localized overheating or energy waste caused by cooling delays. An online parameter self-adjustment mechanism is introduced. The system dynamically corrects key parameters such as event trigger thresholds, control gain, and decision cycles based on real-time energy consumption, event frequency, and control effect. For example, when high-load events occur frequently, the trigger threshold is automatically lowered to enhance sensitivity; when the control response overshoots, the command amplitude is adjusted to maintain stability. This self-learning capability enables the control system to continuously optimize its strategy, maintaining efficient and robust performance in complex and ever-changing environments. A flexible and efficient adaptive dual-timescale cooperative control method for high-performance computing centers is constructed through "event-triggered decision-making, bidirectional cooperative control, and online parameter adaptation." It abandons the rigid control mode of traditional fixed cycles, instead dynamically adjusting the decision-making rhythm and command content according to the actual system state. This significantly reduces overall system energy consumption while ensuring equipment safety and computing efficiency, providing a practical technical path for the green and intelligent operation of large-scale computing infrastructure.

[0103] Figure 5 A fifth alternative flowchart of the data processing method provided in this disclosure embodiment is shown, and will be described according to each step.

[0104] Step S501: Determine the adjustment trigger cycle.

[0105] In some embodiments, the carrier determines the adjustment triggering period based on the minimum decision interval, stability influence coefficient, standard deviation of task-side power, standard deviation of heat dissipation-side power, maximum value of task-side power, and maximum value of heat dissipation-side power. Specifically, this includes:

[0106] in, To minimize the decision interval and ensure the most basic monitoring and response capabilities; This is the stability influence coefficient, used to adjust the magnitude of the influence of stability on the cycle. To implement stability indices for the system based on the standard deviation of task-side power, the standard deviation of heat dissipation-side power, the maximum value of task-side power, and the maximum value of heat dissipation-side power, the calculation methods include:

[0107] in, This represents the standard deviation of the task-side power; optionally, it can be the standard deviation of the task-side normalized power. The standard deviation of the power on the heat dissipation side can optionally be the standard deviation of the normalized power on the heat dissipation side; and Used to describe the degree of fluctuation in the reaction system. The closer the value is to 1, the more stable the system is.

[0108] In step S502, the task side and the heat dissipation side are monitored, and the priority of the event is determined in response to the triggering event.

[0109] In some embodiments, the carrier monitors a first power change rate on the task side and a first temperature change rate on the heat dissipation side. If either change rate exceeds a corresponding threshold, a trigger event is determined. The trigger event indicates that an event exists in the system that causes a change in the monitoring data, and further analysis of the performance change rate, temperature change rate, and duration is needed to determine whether the adjustment trigger conditions are met.

[0110] Specifically, events that trigger the task side can include:

[0111] in, The rate of change of power, The power change threshold; This includes the set of all triggered task-side events; that is, the set of all events where the power change rate is greater than the power change threshold.

[0112] Specifically, events that trigger the heat dissipation side can include:

[0113] in, For the rate of temperature change, The temperature change threshold; It includes the set of all triggered heat dissipation-side events; that is, the set of all events where the rate of temperature change is greater than the temperature change threshold.

[0114] In some embodiments, the carrier identifies the event source based on a first performance change rate and a first temperature change rate. The weight corresponding to the event source is then used. Amplitude weight Duration weight Source of the incident Performance change rate or temperature change rate and duration of the event Determine the priority of events; specifically including:

[0115] in, This is an amplitude measurement function. Depending on the source of the event, this function takes different values ​​to uniformly quantify the severity of the event. For task-related events... ,but ,or This refers to the rate of change of power on the task side. If it's a heat dissipation-side event... ,but ,or This refers to the rate of change of the cooling return water temperature in the heat dissipation system.

[0116] This is the event type coefficient (determined based on the event source), used to distinguish the fundamental source and nature of an event; the weight corresponding to the event source. This is used to control the weight of event type in the overall score; magnitude weight. This is used to control the weight of the magnitude of event changes in the overall score. Duration of the event; This is the duration weight, used to control the weight of event duration in the overall score.

[0117] This phase involves sensing, identifying, and evaluating task-side and heat dissipation-side events (such as "rapid changes in temperature gradient") in real time, determining their "importance" and "urgency," i.e., priority.

[0118] Step S503: Determine whether the adjustment trigger condition is met.

[0119] In some embodiments, the carrier is determined to meet the adjustment triggering condition if it reaches the adjustment triggering period or the time priority is greater than the priority threshold.

[0120] Specifically, the carrier can determine at each time t whether the current time meets the adjustment trigger condition. If it does, step S504 is executed; if it does not, steps S501 to S503 are executed, specifically including:

[0121] When the priority of the event is detected Greater than the priority threshold In the case of determining that the adjustment trigger condition is met; or, when the adjustment trigger period is detected to have arrived. The adjustment triggering condition is determined when the event occurs at an integer multiple of the adjustment triggering period. This combination of two conditions ensures that high-priority events are processed in advance, without waiting for the adjustment triggering period to arrive; furthermore, triggering at integer multiples of the adjustment triggering period prevents low-priority events from being missed.

[0122] Step S504: Determine the task control instructions based on the first performance index of the electronic device.

[0123] In some embodiments, the carrier determines the task control instructions, including:

[0124] In some embodiments, in response to the power included in the first performance metric Greater than the first threshold And the priority of events Greater than the second threshold If the event occurs, the task corresponding to the event will be stopped and suspended. ).

[0125] In other embodiments, in response to the first node load included in the first performance metric Less than the third threshold And the first performance indicator includes the resource utilization rate of the first node. Less than the fourth threshold Then the task corresponding to the event will be scheduled to the first node ( The task is performed by the first node. The first node includes nodes with sufficient cooling capacity and lower current pressure.

[0126] In some other embodiments, for scenarios other than the two cases mentioned above ( The carrier can adjust the execution order of tasks. This allows tasks to be processed earlier, later, or in the current execution order.

[0127] Step S505: Determine the temperature control command based on the predicted performance indicators of the electronic device.

[0128] In some embodiments, the carrier determines the predicted power change from the current moment to a first moment based on historical events, historical tasks, and historical performance indicators of the electronic device; the first moment is a future moment, i.e. a moment that has not yet occurred.

[0129] Specifically, the carrier can train the model based on historical events, historical tasks, historical performance indicators, and power changes of electronic devices collected before the current moment, so that the model has the ability to predict power changes based on historical events, historical tasks, and historical performance indicators of electronic devices.

[0130] In some embodiments, the carrier generates a forward-looking control sequence for the refrigeration system, enabling it to respond in advance to predicted load changes. Specifically, this includes: determining a cooling parameter adjustment sequence from the current moment to a first moment based on predicted power changes, maximum power values, thermal safety scheduling adjustment factors, control gain coefficients, and current temperature control commands; and determining the cooling parameter adjustment sequence from the current moment to the first moment as the temperature control command. The maximum power value can be determined based on predicted power changes or historical performance indicators; the thermal safety scheduling adjustment factor is used to dynamically adjust the control intensity according to the proximity of the current temperature to a safety threshold; the control gain coefficient is used to control the adjustment step size; the cooling parameter adjustment sequence is a forward-looking control sequence for the heat dissipation system, including adjustment strategies from the current moment to the first moment, such as return water temperature and flow rate setpoints. Specifically, this includes:

[0131] in, From the current moment To the first moment of the future The cooling parameter adjustment sequence (such as return water temperature, flow rate setpoint); To predict power changes; It serves as a thermal safety scheduling adjustment factor; This is to control the gain coefficient.

[0132] When a sudden increase in computing tasks (task-side task surge) is predicted, the carrier will instruct the CDU to gradually reduce the return water temperature over the next few minutes, thereby establishing sufficient heat dissipation capacity before the thermal shock occurs and effectively avoiding local overheating and system oscillation.

[0133] Step S506: Adaptively adjust parameters.

[0134] In some embodiments, the carrier may adjust the weights in the formula used to determine event priorities in step S502 based on the system's recent energy efficiency performance:

[0135] in, , for time The overall energy consumption cost; For mission-side power, For heat dissipation side power, As the first coefficient, The second coefficient; If energy consumption costs decrease as a result of performance improvement efforts, then we will continue to optimize in the current direction. To control and adjust the step size; The preset adjustment amount ensures the stability and controllability of the optimization process.

[0136] In other embodiments, the carrier is based on a priority threshold corresponding to the second time step. The number of events that meet the adjustment trigger conditions within the third time interval. Reference value for the expected maximum event frequency within the third time interval and dynamic adjustment coefficient The priority threshold is updated to determine whether the event at the fourth time point meets the adjustment trigger condition. The third time interval is the time interval before the second time point. The dynamic adjustment coefficient is used to automatically increase its value to filter interference and enhance stability when events are too frequent, and conversely, to decrease its value to improve system response sensitivity when events are less frequent. Specifically, it includes:

[0137] in, The number of events that meet the adjustment triggering conditions within the third time interval; This is a reference value for the expected maximum event frequency within the third time interval; This is a dynamic adjustment coefficient used to automatically increase the threshold to filter out interference and enhance stability when events are too frequent; conversely, it lowers the threshold to improve the system's response sensitivity when events are less frequent. For event sequence ( or The parameter is adjusted to reflect the influence of the preceding event on the following event.

[0138] In yet another embodiment, the carrier monitors the power coordination changes of the IT and cooling systems to correct the control gain parameters involved in step S505 in real time. To suppress oscillations and optimize response speed, the following measures are taken:

[0139] in, and These represent the power changes of the cooling system on the heat dissipation side and the IT system on the task side, respectively. This serves as an indicator of synergistic effect. When the two changes in opposite directions, it indicates that the control may be having a counterproductive effect, and the control gain parameter needs to be increased. If the directions are the same, overshoot may occur, and the control gain parameter needs to be reduced. . This is the gain adjustment coefficient.

[0140] In this way, the carrier is transformed into a "smart and efficient dispatching and command" system. It does not pursue a theoretically perfect global optimal solution, but rather achieves near-optimal results in practice at extremely low cost by making the right decisions at the right time.

[0141] Step S507: Update and adjust the trigger cycle.

[0142] In some embodiments, to further reduce control overhead during steady-state operation, when the system remains stable, the adjustment trigger cycle is automatically extended, and an energy-saving silent mode is entered, specifically including:

[0143] in, The duration of stable operation of electronic devices; This is the silent adjustment coefficient, which controls the maximum extension of the decision interval; This is the maximum stable duration, and also a reference value for the maximum sustained stable time, used for normalization.

[0144] Thus, through the data processing method provided in this embodiment, by real-time monitoring of key indicators such as power changes and cooling temperature gradients on the task side, high-priority events (such as sudden increases in tasks) are automatically identified, and collaborative decision-making is immediately triggered. During stable system operation, the decision interval is automatically extended, entering a low-overhead "energy-saving silent mode." This mechanism enables the control system to respond quickly at critical moments and reduce intervention during idle times, achieving a balance between responsiveness and control efficiency. At each collaborative decision-making moment, bidirectional control commands are generated synchronously: precise job scheduling intervention is implemented on the task side, such as delaying submission or guiding tasks to nodes with sufficient cooling capacity; a set of forward-looking parameter adjustment sequences is output for the refrigeration system, such as gradually adjusting the return water temperature and flow rate to respond in advance to predicted load changes. This mechanism achieves a bidirectional closed loop at the action level, effectively suppressing system oscillations and avoiding localized overheating or energy waste caused by cooling delays. An online parameter self-adjustment mechanism is introduced. The system dynamically corrects key parameters such as event trigger thresholds, control gain, and decision cycles based on real-time energy consumption, event frequency, and control effect. For example, when high-load events occur frequently, the trigger threshold is automatically lowered to enhance sensitivity; when the control response overshoots, the command amplitude is adjusted to maintain stability. This self-learning capability enables the control system to continuously optimize its strategy, maintaining efficient and robust performance in complex and ever-changing environments. A flexible and efficient adaptive dual-timescale cooperative control method for high-performance computing centers is constructed through "event-triggered decision-making, bidirectional cooperative control, and online parameter adaptation." It abandons the rigid control mode of traditional fixed cycles, instead dynamically adjusting the decision-making rhythm and command content according to the actual system state. This significantly reduces overall system energy consumption while ensuring equipment safety and computing efficiency, providing a practical technical path for the green and intelligent operation of large-scale computing infrastructure.

[0145] Figure 6 A schematic diagram of an application scenario provided by an embodiment of this disclosure is shown.

[0146] In some embodiments, the carrier may be Figure 6 The adaptive dual-time-scale controller shown is referred to as the controller.

[0147] The controller can realize event-triggered dynamic decision (instruction) generation, bidirectional instruction coordination and active control, as well as parameter adaptation and dynamic threshold adjustment.

[0148] The event-triggered dynamic decision (instruction) generation part includes monitoring test indicators and triggering decisions for high-priority events; bidirectional instruction coordination and proactive control includes synchronizing IT load scheduling on the task side and adjusting cooling parameters of the cooling system on the heat dissipation side; parameter adaptation and dynamic threshold adjustment includes adjusting parameters based on the energy consumption, number of events, efficiency, etc. of electronic devices to continuously optimize the system state of electronic devices.

[0149] In some embodiments, the task side of this disclosure may include an IT high-performance computing server (HPC cluster), and the heat dissipation side may include a liquid cooling distribution unit (CDU).

[0150] like Figure 6 As shown, the controller first monitors the power change rate on the task side and the temperature change rate (cooling water return temperature gradient) on the heat dissipation side in real time. When the detected change exceeds a preset threshold, a corresponding event is triggered. Subsequently, a weighted scoring mechanism is used to comprehensively evaluate the priority of events, considering factors such as event type, change magnitude, and duration. Based on the evaluation results, the decision-making rhythm is dynamically adjusted: high-priority events are immediately triggered with collaborative decisions, while during stable system operation, the decision interval is automatically extended, entering an energy-saving silent mode to reduce control overhead. At each decision moment, bidirectional command coordination is executed synchronously. Scheduling task control commands are generated for the IT system on the task side, including delaying job execution, adjusting task order, or guiding tasks to nodes with better heat dissipation conditions; simultaneously, a forward-looking parameter adjustment sequence is generated for the cooling system on the heat dissipation side, such as adjusting the return water temperature and flow / differential pressure setpoints in stages. This bidirectional coordination ensures the synchronization between IT load changes and cooling response. Finally, an adaptive adjustment mechanism continuously optimizes operating parameters. Based on real-time collected energy consumption performance data, key parameters such as event trigger thresholds, control gain coefficients, and weighting parameters are dynamically adjusted. This closed-loop control process realizes a complete cycle from state perception, event assessment, dynamic decision-making to instruction execution, effectively coordinating the rapidly changing IT load and the slow-response cooling system, and significantly improving the overall energy efficiency ratio while ensuring equipment safety.

[0151] Figure 7 The diagram shown is a scheduling schematic of an embodiment of this disclosure.

[0152] like Figure 7 As shown, the horizontal axis represents task-side scheduling commands with a time scale of seconds; the vertical axis represents cooling-side scheduling commands with a time scale of minutes. The data processing method provided in this embodiment enables dual-scale fast-slow coordinated control.

[0153] To verify the feasibility of the data processing method provided in this embodiment, a simulation environment simulating a supercomputing center was constructed, comprising 6 racks, each rack contributing 10 servers (a total of 120 nodes), and 2 cooling distribution units (CDUs).

[0154] Thermodynamic model: accurately simulated the dynamic relationship between the secondary side supply and return water temperature and flow rate of the CDU and the server inlet temperature.

[0155] Test scenarios: Four typical scenarios were designed to comprehensively evaluate performance: Scenario 1 (Normal workday): Moderate load intensity, outdoor temperature fluctuates between 20-28°C.

[0156] Scenario 2 (Peak Flood): Simulates sudden high-intensity computing demands to test the system's emergency response capabilities.

[0157] Scenario 3 (Low Load at Night): Running under low load, focusing on testing its ability to avoid "overcooling".

[0158] Scenario 4 (Hot Day): When the outdoor temperature rises above 35°C, the cooling efficiency decreases, testing the system's heat dissipation capacity and stability under extreme conditions.

[0159] state space : This includes candidate task requirements, coolant flow rate, IT load status (available cores, utilization, and power consumption of each server), thermal status (CDU return / supply water temperature, flow rate / pressure difference), and ambient temperature.

[0160] Action space : ,in These are discrete job scheduling actions. This is for continuous control commands (return water temperature, flow rate setpoints) for multiple CDUs.

[0161] objective function :

[0162] The results are compared below:

[0163] Among them, task-side optimization includes FIFO first-in-first-out scheduling, and Slurm sets the FIFO scheduling strategy.

[0164] Heat dissipation optimization includes Fixed-PID, which uses a fixed temperature setpoint of CDU and combines it with a PID controller to adjust the flow rate.

[0165] Two-way regulation includes an adaptive dual-timescale control mechanism.

[0166] Data shows that creating a more stable and cooler thermal environment through an adaptive dual-timescale controller provides favorable conditions for the rapid completion of computational tasks, indirectly improving overall computational efficiency. It avoids performance throttling caused by overheating: good thermal management ensures that the server CPU always runs at peak performance. It enables smarter load scheduling: through proactive prediction and bidirectional adjustment, it may reduce job queuing time or allocate jobs to more suitable nodes.

[0167] DTC tuning process pseudocode:

[0168] Figure 8 A schematic diagram of an optional structure of the data processing apparatus provided in an embodiment of this disclosure is shown, and the details will be described in terms of each part.

[0169] In some embodiments, the data processing apparatus includes a judgment unit, a determination unit, and an execution unit.

[0170] The judgment unit is used to determine whether the adjustment triggering condition is met based on the first performance change rate and the first temperature change rate of the electronic device. The determining unit is configured to, in response to the satisfaction of the adjustment triggering condition, determine a task control command based on a first performance index of the electronic device, and simultaneously determine a temperature control command based on a predicted performance index of the electronic device. An execution unit is used to schedule the execution of tasks of electronic devices based on the task control instructions and to control the refrigeration system based on temperature control instructions. The time scale of the task control command is smaller than that of the temperature control command; the time scale includes the time unit required to execute the control command, or the time unit required to respond to the control command.

[0171] In some alternative embodiments, the data processing apparatus may further include an adjustment unit.

[0172] The adjustment unit, after scheduling the task execution of the electronic device based on the task control command and controlling the refrigeration system based on the temperature control command, is also used for one of the following: Obtain a second performance index of the electronic device; based on the second performance index, update the weights of each parameter; the weights of each parameter are used to determine the priority of events caused by the rate of performance change or the rate of temperature change; Obtain the number of events that meet the adjustment triggering conditions; update the priority threshold based on the number of events that meet the triggering conditions, the maximum number of events, and the control gain parameter; the priority threshold is used to determine whether the adjustment triggering conditions are met. Obtain the second performance change rate and the second temperature change rate of the electronic device, and update the control gain parameter based on the second performance change rate and the second temperature change rate.

[0173] The judgment unit is specifically used to confirm that the event corresponding to the current moment meets the adjustment triggering condition in response to the first performance change rate being greater than the performance change threshold or the first temperature change rate being greater than the temperature change threshold.

[0174] The judgment unit is also used to confirm the source of the event based on the first performance change rate and the first temperature change rate; The priority of an event is determined based on its source, rate of performance change or rate of temperature change, duration of the event, and the weight of each parameter. If the priority of the event is greater than the priority threshold, then the event corresponding to the current moment is determined to meet the adjustment triggering condition.

[0175] The determining unit is specifically used for one of the following: If the power included in the first performance metric is greater than a first threshold, and the priority of the event included in the first performance metric is greater than a second threshold, then the task corresponding to the event is stopped. If the first node load, which is included in the first performance metric, is less than the third threshold, and the first node resource utilization, which is included in the first performance metric, is less than the fourth threshold, then the task corresponding to the event is scheduled to the first node, and the first node executes the task. Adjust the execution order of the tasks corresponding to the event.

[0176] The determining unit is specifically used for one of the following: Based on historical events, historical tasks, and historical performance indicators of electronic devices, determine the predicted power change from the current moment to the first moment. Based on the predicted power change, maximum power value, thermal safety scheduling adjustment factor, control gain coefficient, and current temperature control command, determine the cooling parameter adjustment sequence from the current moment to the first moment; The cooling parameter adjustment sequence from the current moment to the first moment is determined as the temperature control command.

[0177] The judgment unit is also configured to, in response to the arrival of the adjustment trigger cycle, determine a task control command based on a first performance index of the electronic device and a temperature control command based on a predicted performance index of the electronic device. The adjustment triggering period is determined based on the minimum decision interval, stability influence coefficient, standard deviation of task-side power, standard deviation of heat dissipation-side power, maximum value of task-side power, and maximum value of heat dissipation-side power.

[0178] The adjustment unit is also used to update the adjustment trigger cycle based on the duration of stable operation of the electronic device, the silent adjustment coefficient, and the maximum stable duration.

[0179] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0180] Figure 9A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0181] like Figure 9 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0182] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0183] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as data processing methods. For example, in some embodiments, the data processing method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the data processing method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform data processing methods by any other suitable means (e.g., by means of firmware).

[0184] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0185] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0186] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0187] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0188] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0189] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0190] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0191] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0192] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A data processing method, the method comprising: Based on the first performance change rate and the first temperature change rate of the electronic device, determine whether the adjustment trigger condition is met; In response to the fulfillment of the adjustment triggering condition, a task control command is determined based on a first performance index of the electronic device, and a temperature control command is determined based on a predicted performance index of the electronic device. The task execution of electronic devices is scheduled based on the task control instructions, and the refrigeration system is controlled based on the temperature regulation instructions. The time scale of the task control command is smaller than that of the temperature control command; the time scale includes the time unit required to execute the control command, or the time unit required to respond to the control command.

2. The method according to claim 1, wherein after scheduling the task execution of the electronic device based on the task control command and controlling the refrigeration system based on the temperature control command, the method further includes one of the following: Obtain a second performance index of the electronic device; based on the second performance index, update the weights of each parameter; the weights of each parameter are used to determine the priority of events caused by the rate of performance change or the rate of temperature change; Obtain the number of events that meet the adjustment trigger conditions; update the priority threshold based on the number of events that meet the trigger conditions, the maximum number of events, and the control gain parameter; The priority threshold is used to determine whether the adjustment trigger condition is met; Obtain the second performance change rate and the second temperature change rate of the electronic device, and update the control gain parameter based on the second performance change rate and the second temperature change rate.

3. The method according to claim 1, wherein determining whether the adjustment trigger condition is met based on the first performance change rate and the first temperature change rate of the electronic device includes: If the first performance change rate is greater than the performance change threshold, or the first temperature change rate is greater than the temperature change threshold, then the event corresponding to the current moment is confirmed to meet the adjustment trigger condition.

4. The method according to claim 1 or 3, further comprising: The source of the event was identified based on the first performance change rate and the first temperature change rate. The priority of an event is determined based on its source, rate of performance change or rate of temperature change, duration of the event, and the weight of each parameter. If the priority of the event is greater than the priority threshold, then the event corresponding to the current moment is determined to meet the adjustment triggering condition.

5. The method according to claim 1, wherein determining the task control instruction based on a first performance indicator of the electronic device includes one of the following: If the power included in the first performance metric is greater than a first threshold, and the priority of the event included in the first performance metric is greater than a second threshold, then the task corresponding to the event is stopped. If the first node load, which is included in the first performance metric, is less than the third threshold, and the first node resource utilization rate, which is included in the first performance metric, is less than the fourth threshold, then the task corresponding to the event is scheduled to the first node, and the first node executes the task. Adjust the execution order of the tasks corresponding to the event.

6. The method according to claim 1, wherein determining the temperature control command based on the predictive performance index of the electronic device includes one of the following: Based on historical events, historical tasks, and historical performance indicators of electronic devices, determine the predicted power change from the current moment to the first moment. Based on the predicted power change, maximum power value, thermal safety scheduling adjustment factor, control gain coefficient, and current temperature control command, determine the cooling parameter adjustment sequence from the current moment to the first moment; The cooling parameter adjustment sequence from the current moment to the first moment is determined as the temperature control command.

7. The method according to claim 1, further comprising: In response to the arrival of the adjustment trigger cycle, a task control command is determined based on the first performance index of the electronic device, and a temperature control command is determined based on the predicted performance index of the electronic device. The adjustment triggering period is determined based on the minimum decision interval, stability influence coefficient, standard deviation of task-side power, standard deviation of heat dissipation-side power, maximum value of task-side power, and maximum value of heat dissipation-side power.

8. The method according to claim 7, further comprising: The adjustment trigger cycle is updated based on the duration of stable operation of the electronic device, the silent adjustment coefficient, and the maximum stable duration.

9. A data processing apparatus, the apparatus comprising: The judgment unit is used to determine whether the adjustment triggering condition is met based on the first performance change rate and the first temperature change rate of the electronic device. The determining unit is configured to, in response to the satisfaction of the adjustment triggering condition, determine a task control command based on a first performance index of the electronic device, and simultaneously determine a temperature control command based on a predicted performance index of the electronic device. An execution unit is used to schedule the execution of tasks of electronic devices based on the task control instructions and to control the refrigeration system based on temperature control instructions. The time scale of the task control command is smaller than that of the temperature control command; the time scale includes the time unit required to execute the control command, or the time unit required to respond to the control command.

10. An electronic device, comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

Images