Power supply control method and device of server and electronic equipment

Abstract

The invention discloses a power supply control method and device of a server and electronic equipment, and relates to the technical field of computers, power supply modes of a group of power supplies comprise a static redundant power supply mode and a dynamic redundant power supply mode, and the method comprises the following steps: obtaining power consumption data of the group of power supplies in a specified historical period to obtain historical power consumption data; on the basis of the historical power consumption data, the load trend of the group of power supplies in a specified future time period is predicted, the predicted load trend of the group of power supplies is obtained, and the specified future time period is a future time period with the current moment or a future moment after the current moment as a starting moment; and according to the predicted load trend and the specified mode switching condition, dynamically switching the power supply mode of a group of power supplies, thereby solving the technical problems of low efficiency of the power supplies and serious electric energy waste in a power supply control method in the related technology, and achieving the technical effects of improving the efficiency of the power supplies and reducing the electric energy waste.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a power supply control method, device, and electronic device for a server. Background Art

[0002] In related technologies, servers typically use a 1+1 or N+1 redundant power supply design to ensure power supply reliability. The redundant power supplies will be in standby mode at full power, resulting in unnecessary energy waste. In addition, server power supplies are typically configured according to peak load and maintain high power supply during actual low-load business operation.

[0003] Furthermore, the redundant power supply design described above often results in some power supplies operating at low or no load for extended periods of time. Power supply efficiency typically drops significantly below 30% of their operating load, resulting in low power supply efficiency. This demonstrates that the power supply control methods used in related technologies suffer from low power supply efficiency and significant energy waste. Summary of the Invention

[0004] The present application provides a power supply control method, device and electronic device for a server, so as to at least solve the technical problems of low power supply efficiency and serious energy waste in the power supply control method in the related art.

[0005] The present application provides a power supply control method for a server, wherein the server has a group of power supplies, and the power supply modes of the group of power supplies include a static redundant power supply mode and a dynamic redundant power supply mode. In the static redundant power supply mode, the number of redundant power supplies in the group of power supplies is fixed, and when not enabled, the redundant power supplies are in a standby state; in the dynamic redundant power supply mode, the number of redundant power supplies in the group of power supplies is dynamically adjusted according to the load of the group of power supplies, and when not enabled, the redundant power supplies are in a shutdown state. The method comprises: obtaining power consumption data of the group of power supplies within a specified historical period to obtain historical power consumption data, wherein the specified historical period is a historical period with a current moment or a historical moment before the current moment as an end time; predicting a load trend of the group of power supplies within a specified future period based on the historical power consumption data to obtain a predicted load trend of the group of power supplies, wherein the specified future period is a future period with a future moment after the current moment as a start time; and dynamically switching the power supply mode of the group of power supplies according to the predicted load trend and a specified mode switching condition.

[0006] The present application also provides a power supply control device for a server. The server has a set of power supply units, and the power supply modes of the set of power supply units include a static redundant power supply mode and a dynamic redundant power supply mode. Among them, in the static redundant power supply mode, the number of redundant power supply units in the set of power supply units is fixed, and the redundant power supply units are in a standby state when not enabled; in the dynamic redundant power supply mode, the number of the redundant power supply units in the set of power supply units is dynamically adjusted according to the load of the set of power supply units, and the redundant power supply units are in an off state when not started; the device includes: an acquisition unit, configured to acquire the power consumption data of the set of power supply units in a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period ending at the current moment or a historical moment before the current moment; a prediction unit, configured to predict the load trend of the set of power supply units in a specified future period based on the historical power consumption data to obtain the predicted load trend of the set of power supply units, where the specified future period is a future period starting at the current moment or a future moment after the current moment; a switching unit, configured to dynamically switch the power supply mode of the set of power supply units according to the predicted load trend and specified mode switching conditions.

[0007] The present application also provides an electronic device, including: a memory, configured to store a computer program; a processor, configured to implement the steps of any one of the above server power supply control methods when executing the computer program.

[0008] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above server power supply control methods are implemented.

[0009] The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of any one of the above server power supply control methods are implemented.

[0010] With this application, a group of power supplies has a static redundant power supply mode and a dynamic redundant power supply mode. The static redundant power supply mode always keeps the redundant power supply in a standby state, which is suitable for high-load scenarios. The dynamic redundant power supply module can adjust the number of power supplies in the working state according to the actual load, which is suitable for low-load scenarios. Since the future load trend of the power supply can be predicted and the power supply mode of a group of power supplies can be dynamically switched based on the load trend, redundant power supplies can be reduced in low-load scenarios, reducing power loss. Therefore, the technical problems of low efficiency of the power supply and serious power waste existing in the power supply control method in the related art can be solved, and the technical effects of improving the efficiency of the power supply and reducing power waste can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] To more clearly illustrate the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 FIG. is a schematic diagram of an application scenario of a power supply control method for a server according to an embodiment of the present application.

[0013] Figure 2 FIG. is a schematic flowchart of an optional power supply control method for a server according to an embodiment of the present application.

[0014] Figure 3 FIG. is a schematic diagram of an optional power supply control method for a server according to an embodiment of the present application.

[0015] Figure 4 FIG. is a schematic structural diagram of an optional power supply control system for a server according to an embodiment of the present application.

[0016] Figure 5 FIG. is a structural block diagram of an optional power supply control device for a server according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present application.

[0018] It should be noted that in the description of this application, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or also includes elements inherent to such a process, method, article or device. The terms "first", "second", etc. in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0019] In order to enable those skilled in the art of this technology to better understand the solution of this application, the following further detailed description of this application will be given in conjunction with the drawings and specific implementation manners.

[0020] According to one aspect of the embodiments of this application, a power supply control method for a server is provided. Optionally, in this embodiment, the power supply control method for the above-mentioned server may but is not limited to be applied to a hardware environment including a power supply 102 and a server device 104 as Figure 1 shown. The power supply 102 can be connected to the server device 104 through a power supply interface and can be used to supply power to the server device 104.

[0021] The above-mentioned power supply interface may include but is not limited to at least one of the following: direct power line connection, internal hardware interface connection. The power supply 102 may but is not limited to be a high-voltage direct current power supply, a polyphase power supply, etc. The server device 104 may include but is not limited to a central processing unit (abbreviated as CPU), a graphics processing unit (abbreviated as GPU), a storage device, a memory (abbreviated as RAM), etc.

[0022] The power supply control method for the server in the embodiments of this application can be executed by the server. Figure 2 is a schematic flowchart of an optional power supply control method for a server according to the embodiments of this application. As Figure 2 shown, the process of this method may include the following steps:

[0023]

[0024] Step S202, obtain power consumption data of a group of power supplies within a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period ending at the current moment or a historical moment before the current moment.

[0024] Step S204, predict the load trend of a group of power supplies within a specified future period based on historical power consumption data, so as to obtain the predicted load trend of the group of power supplies, where the specified future period is a future period starting from the current moment or a future moment after the current moment.

[0025] Step S206, dynamically switch the power supply mode of a group of power supplies according to the predicted load trend and the specified mode switching condition.

[0026] Among them, the server has a group of power supplies, and the power supply modes of the group of power supplies include a static redundancy power supply mode and a dynamic redundancy power supply mode. Among them, in the static redundancy power supply mode, the number of redundant power supplies in the group of power supplies is fixed, and when not enabled, the redundant power supplies are in a standby state; in the dynamic redundancy power supply mode, the number of redundant power supplies in the group of power supplies is dynamically adjusted according to the load of the group of power supplies, and when not started, the redundant power supplies are in an off state.

[0027] The power supply control method of the server in this embodiment can be applied to the computer field and applied to scenarios of managing and controlling the power supplies of servers (such as high-efficiency data centers, cloud computing platforms, or edge computing nodes, etc.). A power supply unit (PSU for short) is a type of power supply for power conversion (different from battery-powered power supplies), which is responsible for converting standard alternating current into low-voltage stable direct current for other components that need power supply.

[0028] In the related art, servers usually adopt a 1+1 or N+1 redundant power supply design to ensure power supply reliability, that is, an additional redundant power supply is reserved outside the power supplies required for normal power supply, so as to ensure that the system can still operate normally in the event of a failure of any power supply. However, the above redundant structure will cause some power supplies to run at low load or no load for a long time most of the time, resulting in low efficiency and increased energy consumption. In addition, the power supplies of servers are usually configured according to the peak load, and high-power power supply is still maintained during the low-load operation of actual services, further exacerbating the problem of energy consumption waste.

[0029] Taking a data center (usually deployed with a server cluster) as an example, the power consumption of the data center mainly comes from Information Technology (IT) devices and air conditioners for cooling. To solve the problem of energy consumption waste, it is necessary to comprehensively consider the heat generation and power consumption generated by the operation of the above-mentioned devices. In the related art, although some servers support dynamic fan speed regulation or load-aware adjustment, there is a lack of comprehensive consideration of real-time load, historical operation trends, and business model prediction on the power supply side.

[0030] In order to at least partially solve the above technical problems, in this embodiment, the power supply has at least two power supply modes, and can flexibly adjust the power supply mode according to the predicted load trend, thereby reducing the waste of energy consumption.

[0031] The above two power supply modes include a static redundant power supply mode and a dynamic redundant power supply mode. Among them, the static redundant power supply mode is similar to the 1+1 or N+1 redundant power supply mode in the related art. At least one redundant standby power supply is reserved outside N actual load power supplies. The redundant power supply is in a standby state before being enabled. For example, if 3 PSUs are required for the normal operation of the system, 4 PSUs can be deployed, and one of them is a standby redundant PSU. The dynamic redundant power supply mode can dynamically determine the number of redundant power supplies according to the real-time load. For example, when the load is continuously lower than a load threshold, the redundant PSU can be turned off to reduce energy consumption. The characteristic of the static redundant power supply mode is that the redundant resources are always in the power-on standby state, which can ensure that the system does not crash when any one of the power supplies fails. This mode has a high energy consumption, and the redundant power supplies will continue to consume energy even without load; the characteristic of the dynamic redundant power supply mode is that the power supplies can be dynamically turned off and on (for example, it can be switched from 4 PSUs to 3 or 2 online), which can save unnecessary redundant standby power consumption, but precise prediction and fast response capabilities are required to ensure the stable operation of the power supply function.

[0032] In order to achieve flexible and accurate switching between the above two modes, in this embodiment, it is necessary to obtain historical power consumption data and predict future load trends based on this. Here, obtaining historical power consumption data means obtaining the power consumption data of the set of power supplies within a specified historical period to obtain historical power consumption data.

[0033] Optionally, the historical power consumption data can be obtained by using a Baseboard Management Controller (BMC) or a Power Management Bus (PMBus). The historical power consumption data can include power input current, voltage, power, etc. In addition, the historical power consumption data can be collected according to a specified sampling period, and the specified sampling period can be set according to experience. For example, it can be 1 second, 1.5 seconds, or other values, which are not limited in this embodiment.

[0034] After obtaining the historical power consumption data, the load trend of a group of power supplies within a specified future period can be predicted based on the historical power consumption data, and the predicted load trend of the group of power supplies can be obtained. Here, the specified future period is a future period starting from the current moment or a future moment after the current moment. For example, the specified future period can be the next five minutes from the current moment. Optionally, the load trend can be predicted at a certain period, and this period should be less than the range of the specified future period. For example, the load trend for the next five minutes can be predicted every three minutes.

[0035] Optionally, the load trend can be predicted by a pre-trained deep learning-based load trend prediction model, such as a Long Short-Term Memory (LSTM) network, a Gated Recurrent Unit (GRU) network, or a Transformer model, etc. The input of the above models is the historical power consumption data. To improve the accuracy of load trend prediction, a sensor gateway can also be connected to perform normalized modeling on the power consumption of server devices such as CPUs, memories, and GPUs, and use it as the input of the model as well.

[0036] After obtaining the predicted load trend, the power supply mode of a group of power supplies can be dynamically switched according to the predicted load trend and the specified mode switching conditions. For example, when it is predicted that the load is high within the specified future period, it can be switched to the static redundant power supply mode to ensure stable power supply; when it is predicted that the load is low within the specified future period, it can be switched to the dynamic redundant power supply mode to improve the energy efficiency ratio and reduce waste.

[0037] It should be noted that the power supply efficiency of the power supply unit usually drops significantly when the working load is below 30% of its rated value. Keeping the power supply unit at a relatively high load can reduce waste. In this embodiment, in the case of a low load, such as during the low-traffic period at night, the power supply mode can be switched to the dynamic redundant power supply mode, which can reduce the number of power supply units in operation, thereby increasing the load of each power supply unit in operation and improving the power supply efficiency. If the static redundant power supply mode is still used during the low-traffic period at night, not only will the redundant power supply units be on standby at full power, but the load of each power supply unit in operation will also be low, resulting in a large amount of unnecessary power waste. Therefore, dynamically adjusting the power supply mode can effectively reduce power waste and improve the efficiency of the power supply unit.

[0038] Through the embodiments provided in this application, the server has a set of power supply units, and the power supply modes of the set of power supply units include the static redundant power supply mode and the dynamic redundant power supply mode. Among them, in the static redundant power supply mode, the number of redundant power supply units in the set of power supply units is fixed, and the redundant power supply units are in a standby state when not enabled; in the dynamic redundant power supply mode, the number of redundant power supply units in the set of power supply units is dynamically adjusted according to the load of the set of power supply units, and the redundant power supply units are in an off state when not started. The method includes: obtaining the power consumption data of the set of power supply units during a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period ending at the current time or a historical time before the current time; predicting the load trend of the set of power supply units during a specified future period based on the historical power consumption data to obtain the predicted load trend of the set of power supply units, where the specified future period is a future period starting at the current time or a future time after the current time; dynamically switching the power supply mode of the set of power supply units according to the predicted load trend and the specified mode switching condition, which solves the technical problems of low efficiency of the power supply unit and serious power waste existing in the power supply control method in the related art, improves the efficiency of the power supply unit, and reduces power waste.

[0039] In an exemplary embodiment, the specified mode switching condition includes a first switching condition for switching from the static redundant power supply mode to the dynamic redundant power supply mode. The first switching condition includes that the duration for which the load of the set of power supply units is less than the first load threshold is greater than or equal to the specified time threshold. Dynamically switching the power supply mode of the set of power supply units according to the predicted load trend and the specified mode switching condition includes: when the power supply mode of the set of power supply units is the static redundant power supply mode and the predicted load trend indicates that the predicted load of the set of power supply units during the specified future period is less than the first load threshold for a duration greater than or equal to the specified time threshold, switching the power supply mode of the set of power supply units to the dynamic redundant power supply mode.

[0040] In the static redundant power supply mode, even when the server is in a low-load or even idle state, such as low-load operation at night or during off-peak periods, the redundant power supply units will remain on standby continuously and consume electrical energy continuously. To solve the problem of excessive power consumption of redundant power supply units under low load, the power supply mode can be switched to the dynamic redundant power supply mode when the first switching condition is met, so as to avoid unnecessary power consumption of redundant power supply units, and increase the load of each working power supply unit, thereby improving the power supply efficiency of the power supply unit.

[0041] In this embodiment, the specified mode switching conditions include the first switching condition for switching from the static redundant power supply mode to the dynamic redundant power supply mode. The first switching condition includes that the predicted load trend indicates that the duration for which the predicted load of a group of power supply units is less than the first load threshold within a specified future period is greater than or equal to the specified time threshold, that is, it is predicted that the load of a group of power supply units remains low. Here, both the first load threshold and the specified time threshold can be set according to experience. For example, the first load threshold can be 50%, 30% or other values, and the specified time threshold needs to be less than the predicted specified future period. For example, when the specified future period is the next 30 minutes, the specified time threshold can be 15 minutes, 20 minutes or other values less than 30 minutes. This embodiment does not make any limitations in this regard.

[0042] Optionally, before switching to the dynamic redundant power supply mode, a redundant safety assessment can also be performed. If the redundant safety assessment fails, the mode will not be switched to the dynamic redundant power supply mode. Here, the conditions for passing the redundant safety assessment can include: the existing power supply units support 150% peak capacity of the load; there is margin in both the Power Distribution Unit (PDU) and the Uninterruptible Power Supply (UPS); the current time is not a critical period such as system maintenance, upgrade, or switching; and indicators such as the current device temperature and safety voltage are normal. After passing the above redundant safety assessment, one or more redundant power supply units can be turned off by means of software power-off or physical power-off.

[0043] Optionally, the method of performing power supply mode switching may include hardware layer methods. For example, through PSU / PDU control, protocols such as PMBus, Redfish, Intelligent Platform Management Interface (IPMI for short) are used to implement the power-on / off control of the PSU. For example, in the case of using the IPMI protocol, the IPMI command can be "chassis power off <PSU_id>". In addition, the input / output current, voltage, power, etc. of the PSU can be read in real time through hardware layer methods to judge the power supply path status and achieve continuous monitoring of the PSU status.

[0044] Optionally, the method of performing power supply mode switching may also include software layer policy scheduling. For example, first, the corresponding module in the system proposes a redundancy reduction suggestion, and the central management control unit decides whether to perform the redundancy reduction operation. Before confirming the execution of the redundancy reduction operation, the feasibility calculated by the "minimum redundancy margin" formula needs to be satisfied, and the switching operation record is written into the log after the power supply mode switching operation is executed. Here, the code for calculating the "minimum redundancy margin" formula can be:

[0045] Ptotal_available≥Ppredicted_peak×safety_marginP_{\text{total\_available}}\geq P_{\text{predicted\_peak}}\times

[0046] \text{safety\_margin}Ptotal_available≥Ppredicted_peak×safety_margin

[0047] Through this embodiment, when it is predicted that the load trend remains low, switching to the dynamic redundant power supply mode can avoid the inefficient operation of the redundant power supply unit, reduce unnecessary power consumption; in addition, reducing the unnecessary operation of the redundant power supply unit can also extend the service life of the power supply unit.

[0048] In an exemplary embodiment, the specified mode switching condition includes a second switching condition for switching from a dynamic redundant power supply mode to a static redundant power supply mode. The second switching condition includes that the load of a group of power supply units is greater than or equal to a second load threshold, where the second load threshold is greater than the first load threshold. According to the predicted load trend and the specified mode switching condition, dynamically switching the power supply mode of a group of power supply units includes: when the power supply mode of a group of power supply units is the static redundant power supply mode and the predicted load trend indicates that there is a future moment when the predicted load of a group of power supply units is greater than or equal to the second load threshold within a specified future period, switching the power supply mode of a group of power supply units to the static redundant power supply mode.

[0049] In the dynamic redundant power supply mode, the unactivated redundant power supply units are in the off state instead of the standby state. When the server faces high load, sudden load increase, or failure of some power supply units, the redundant power supply units in the off state cannot enter the working state in time, which may lead to problems such as power shortage or power overload. To solve the above problems, the power supply mode can be switched to the static redundant power supply mode when the second switching condition is met, keeping a certain number of redundant power supply units in the standby state to ensure that they can enter the working state in time when needed and guarantee the continuity and stability of power supply.

[0050] In this embodiment, the specified mode switching condition includes a second switching condition for switching from a dynamic redundant power supply mode to a static redundant power supply mode. The second switching condition includes that the predicted load trend indicates that there is a future moment when the predicted load of a group of power supply units is greater than or equal to the second load threshold within a specified future period, that is, a moment with a relatively high load is predicted. Here, the second load threshold can be set according to experience and needs to be greater than the first load threshold. For example, when the first load threshold is 50%, the second load threshold can be 70%, 75%, or other values greater than the first load threshold, which is not limited in this embodiment.

[0051] Optionally, the conditions for switching to the static redundant power supply mode may further include: abnormal voltage / current or fluctuations in any power supply path; the system enters the high-availability or service go-live stage; the device operating temperature rises above a certain threshold (indicating a decrease in power efficiency). When any of the above conditions is met, the system can preheat or wake up the redundant power supply units and switch to the static redundant power supply mode. Here, the method of switching the power supply mode is similar to that in the foregoing embodiments and will not be elaborated here.

[0052] In this embodiment, when a high-load trend is predicted, the system switches to the static redundant power supply mode, increasing the redundancy of the power supply in a timely manner before the load increases, effectively avoiding power supply shortages or power overloads, thereby ensuring the continuity of services and the stability of server operation.

[0053] In an exemplary embodiment, the load trend of a set of power supplies within a specified future period is predicted based on historical power consumption data, obtaining the predicted load trend of the set of power supplies, including: extracting features from the historical power consumption data to obtain a historical power consumption sequence; inputting the historical power consumption sequence, or the historical power consumption sequence and at least one of the following features into a load prediction model to obtain the predicted load trend output by the load prediction model: the historical power consumption load sequence of the set of power supplies within a specified historical period, the historical utilization rate sequence of the server's power consumption within a specified historical period, and the task queue characteristics processed by the server's power consumption within a specified historical period, where the load prediction model is a time series model for load prediction.

[0054] In this embodiment, to accurately predict the predicted load trend, the load trend within a specified future period is predicted through a load prediction model, where the load prediction model is a time series model for load prediction.

[0055] Optionally, the load prediction model can adopt a two-layer LSTM neural network structure. The two-layer LSTM neural network structure is a deep learning model composed of two stacked LSTM layers. Each layer contains independent gating mechanisms (forget gate, input gate, output gate) and memory units, which can be used for the prediction and analysis of time series data, and has the advantages of processing long sequence dependencies and capturing complex patterns. In this embodiment, it can be used to accurately predict the load trend of the power supply.

[0056] The input of the load prediction model includes the historical power consumption sequence of the power supply, the historical power consumption load sequence of the power supply, the historical utilization rate sequence of the server, and the historical task queue characteristics of the server. Among them, the historical power consumption sequence of the power supply is obtained by extracting features from the acquired historical power consumption data of the power supply, and can reflect information such as the average power consumption of the power supply within a specified historical period; the historical power consumption load sequence of the power supply is a set of historical power consumption load sequences of the power supply within a specified historical period, and can reflect information such as the load peak and valley values of the power supply within a specified historical period; the historical utilization rate sequence of the server is the historical utilization rate sequence of the power consumption of the server within a specified historical period, and can reflect information such as the average utilization rate, utilization rate peak, and utilization rate valley value of the server within a specified historical period; the historical task queue characteristics of the server are the task queue characteristics processed by the power consumption of the server within a specified historical period, and based on the length of the task queue, the amount of tasks processed by the server within a specified historical period can be reflected. Based on the above input information, the predicted load trend of a set of power supplies within a specified future period can be predicted according to the regression result of the load prediction model. After testing, the error of the predicted load trend is controlled within ±5%.

[0057] Through this embodiment, by predicting the predicted load trend within a specified future period through the load prediction model, and using inputs including multi-dimensional features such as real-time power consumption, historical load, and task queue length, the accuracy of the prediction can be improved, providing a good data basis for dynamically adjusting the power supply mode.

[0058] In an exemplary embodiment, after predicting the load trend of a set of power supplies within a specified future period based on historical power consumption data, the above method further includes: when the power supply mode of a set of power supplies is a dynamic redundant power supply mode, when the predicted load trend of a set of power supplies indicates that the predicted load of a set of power supplies is less than or equal to the third load threshold, start the specified main power supply in a set of power supplies, where other power supplies in a set of power supplies except the main power supply are in a closed state as redundant power supplies; when the predicted load trend of a set of power supplies indicates that the predicted load of a set of power supplies gradually increases from less than or equal to the third load threshold, at a specified moment before the predicted load of a set of power supplies reaches the load threshold in a set of load thresholds, start the corresponding number of redundant power supplies corresponding to the load threshold in a set of load thresholds, where the time interval between the specified moment and the predicted moment when the predicted load of a set of power supplies reaches the load threshold in a set of load thresholds is the specified time interval.

[0059] In the dynamic redundant power supply mode, the number of redundant power supply units can be dynamically adjusted according to the predicted load trend. When the predicted load is low, that is, the predicted load trend of a group of power supply units indicates that the predicted load of a group of power supply units is less than or equal to the third load threshold, only one power supply unit in a group of power supply units can be retained as the main power supply manager, and the main power supply unit continues to work for power supply, while the remaining power supply units are all turned off as redundant power supply units. Here, the third load threshold can be set according to experience. For example, it can be 30%, 25% or other values, and this is not limited in this embodiment. Optionally, the redundant power supply units that are turned off can be in a standby state, the fans stop rotating, and the standby power consumption is less than 3 watts.

[0060] When the predicted load is gradually rising, that is, the predicted load trend of a group of power supply units indicates that the predicted load of a group of power supply units gradually increases from less than or equal to the third load threshold, multiple redundant power supply units can be started before the load reaches the load threshold. Here, the number of started redundant power supply units corresponds to the load threshold in a group of load thresholds, and the higher the load threshold, the more the number of started units. For example, a group of load thresholds can include 60%, 70% and 80%. Among them, 60% can correspond to 2 redundant power supply units, 70% can correspond to 3 redundant power supply units, and 80% can correspond to 4 redundant power supply units. When it is predicted that the predicted load will reach any one of a group of load thresholds, the corresponding number of redundant power supply units can be cold-started at a specified moment before reaching the load threshold. The time interval between the specified moment and the predicted moment when the predicted load of a group of power supply units reaches the load threshold in a group of load thresholds is the specified time interval. Here, the specified time interval can be set according to experience. For example, 30 seconds, 40 seconds or other values. The specified time interval needs to be greater than the startup time of the power supply unit to ensure that the power supply unit can complete startup before the load reaches the load threshold.

[0061] Optionally, precise control of turning off and turning on the power supply unit can be achieved through IPMI or the Inter-Integrated Circuit (I2C) bus. The above two communication protocols can provide an efficient and reliable way for power supply unit management, enabling the server to intelligently adjust the working state of each power supply unit according to load prediction and redundancy strategy.

[0062] Through this embodiment, turning off redundant power supply units in the case of low load and starting redundant power supply units in advance in the case of high load can achieve a rapid response to the load situation, which can not only reduce power consumption but also ensure power supply stability under high load.

[0063] In an exemplary embodiment, the above method further includes: determining a specified load ratio of a power supply in a group of power supplies based on a specified power supply efficiency curve, where the specified power supply efficiency curve is used to indicate the energy conversion efficiency of the power supply in the group of power supplies at a load ratio in a group of load ratios, and the specified load ratio corresponds to the highest energy conversion efficiency in the specified power supply efficiency curve; controlling the power supplies in the group of power supplies that are in the working state to supply power according to the specified load ratio.

[0064] The energy conversion efficiency of the power supply will decrease when the load is too high or too low. To improve the energy conversion efficiency of the power supply, the specified load ratio of the power supply in a group of power supplies can be determined based on the specified power supply efficiency curve, and the power supplies in the working state can be controlled to supply power according to the specified load ratio.

[0065] Here, the specified power supply efficiency curve is used to indicate the energy conversion efficiency of the power supply in the group of power supplies at a load ratio in a group of load ratios, and the specified load ratio corresponds to the highest energy conversion efficiency in the specified power supply efficiency curve. The energy conversion efficiency refers to the energy conversion efficiency when the power supply converts alternating current into direct current for the server device to use, that is, the ratio of the output power to the input power. Optionally, the specified power supply efficiency curve can be the 80 PLUS efficiency curve. The 80 PLUS power standard defines the efficiency indicators of the power supply at 20%, 50%, and 100% loads. The efficiency is the highest at about 50% load. That is, 50% is the specified load ratio, and the energy conversion efficiency at 50% load is the highest energy conversion efficiency in the specified power supply efficiency curve. In this embodiment, the power supplies in the working state can be controlled to operate at a load close to 50% to ensure the high energy conversion efficiency of the power supply.

[0066] Through this embodiment, the power supply is accurately controlled to operate at the load with the highest energy conversion efficiency, so that the power supply is in the high-efficiency area, which can improve the energy conversion efficiency of the power supply in the working state, thereby reducing the waste of electric energy.

[0067] In an exemplary embodiment, after inputting a historical power consumption sequence, or features of a historical power consumption sequence and at least one of the following into a load prediction model, the above method further includes: calculating a set of energy efficiency metrics for the server based on the historical power consumption data, where the set of energy efficiency metrics includes at least one of the following: power usage effectiveness, energy efficiency ratio of cooling; generating a set of candidate operating modes based on the set of energy efficiency metrics for the server, where one of the candidate operating modes in the set of candidate operating modes is used to indicate the redundancy configuration method of a set of power supplies, the number of power supplies in the working state in the set of power supplies, and the load distribution method of the power supplies in the working state in the set of power supplies; determining a target operating mode from the set of candidate operating modes based on the predicted load trend, and performing operating control on the set of power supplies according to the target operating mode.

[0068] The set of energy efficiency metrics may include power usage effectiveness (abbreviated as PUE) and energy efficiency ratio of cooling (abbreviated as EER). Among them, PUE refers to the ratio of the total energy consumption of the data center to the energy consumption of IT equipment. Its value is ≥1, and the closer it is to 1, the lower the energy consumption of non-IT equipment (such as cooling and power supply), and the higher the energy efficiency of the data center; EER refers to the ratio of the cooling capacity of the cooling system to the input power. The higher its value, the higher the cooling efficiency. In order to reduce energy waste, it is necessary to reduce PUE and increase EER, thereby improving the energy utilization efficiency. In this embodiment, based on the PUE and EER of the server, a suitable target operating mode can be selected from the set of candidate operating modes, and the set of power supplies can be controlled according to the target operating mode to optimize the set of energy efficiency metrics.

[0069] In this embodiment, one of the candidate operating modes can be used to indicate the redundancy configuration method of a set of power supplies, the number of power supplies in the working state in the set of power supplies, and the load distribution method of the power supplies in the working state in the set of power supplies. Among them, the redundancy configuration method may include redundancy levels of N, N+1, or N+N, that is, not starting redundant power supplies, starting one redundant power supply, or starting multiple redundant power supplies; the number of power supplies in the working state in the set of power supplies indicates the number of power supplies in the working state and participating in power supply; the load distribution method can indicate whether the power supplies in the working state are powered in a uniform distribution or priority power supply manner. Optionally, in addition to the above three dimensions, one of the candidate operating modes may further include a dynamic adjustment strategy for indicating whether the power supply automatically switches the operating mode according to the predicted load and real-time PUE / EER. One of the candidate operating modes is a combination of the above four dimensions.

[0070] For example, a set of candidate operating modes may include the following three candidate operating modes: Mode A, a high-redundancy and high-power-consumption mode, with a redundancy configuration of 1+1. The number of power supplies in the working state is 4, and the load ratio of each power supply is 37.5%. This mode has low efficiency and a high PUE; Mode B, an optimal efficiency mode, with a redundancy configuration of N (redundant power supplies are not started). The number of power supplies in the working state is 2, and the load ratio of each power supply is 75%. This mode has a high conversion efficiency, the lowest system power consumption, and can maximize the EER; Mode C, a critical balanced redundancy mode, with a redundancy configuration of N+1. The number of power supplies in the working state is 3, and the load ratio of each power supply is 50%. This mode balances redundancy and efficiency.

[0071] After obtaining a set of candidate operating modes, the most suitable target operating mode can be selected from them based on the predicted load trend and executed to ensure that while meeting the business requirements, the power supplies can operate in the most efficient manner.

[0072] Through this embodiment, the current energy efficiency index is calculated in real time, and the target operating mode is determined based on the PUE, EER, and the predicted load trend, which can reduce energy waste while ensuring meeting the business requirements and maintain a good cooling effect.

[0073] In an exemplary embodiment, based on the configuration information, a set of service level agreement rules and a set of alarm levels are determined. Among them, a set of service level agreement rules corresponds to multiple nodes powered by a set of power supplies, indicating the minimum number of power supplies in the standby state, the minimum reserved power supply number, and the alarm threshold corresponding to the multiple nodes. A set of alarm levels includes multiple alarm trigger mechanisms and the alarm handling methods corresponding to the multiple alarm trigger mechanisms; when an alarm trigger mechanism in a set of alarm levels is satisfied, the alarm handling method corresponding to the alarm trigger mechanism is executed.

[0074] Here, the service level agreement (SLA) rule refers to an agreement or contract jointly recognized by both parties between an enterprise providing services and its customers regarding aspects such as the quality, level, and performance of the service. The operation of the server needs to comply with the SLA rules. The SLA rules can correspond to multiple nodes powered by a set of power supplies, indicating the minimum number of power supplies in the standby state, the minimum reserved power supply number, and the alarm threshold corresponding to the multiple nodes. For example, Figure 3As shown, the node types can include GPU nodes, storage nodes, web server nodes, and management control nodes. Among them, the SLA level of GPU nodes is 1, the redundancy requirement is permanent 1+1, the alarm threshold is power drop >5%, the dynamic redundant power supply mode is prohibited, and the minimum number of reserved PSUs is 2; the SLA level of storage nodes is 2, the redundancy requirement is at least N, the alarm threshold is power drop >10%, the dynamic redundant power supply mode is allowed, and the minimum number of reserved PSUs is 1; the SLA level of web server nodes is 2, the redundancy requirement is N+1 and elastic adjustment is allowed, the alarm threshold is power drop >7%, the dynamic redundant power supply mode within ten minutes is allowed, and the minimum number of reserved PSUs is 1; the SLA level of management control nodes is 1, the redundancy requirement is permanent 1+1, the alarm threshold is power drop >3%, the dynamic redundant power supply mode is prohibited, and the minimum number of reserved PSUs is 2. It should be noted that the switching between the aforementioned static redundant power supply mode and dynamic redundant power supply mode also needs to be based on the above SLA rules and can only be switched on the premise of meeting the SLA rules.

[0075] When one of the alarm trigger mechanisms in a set of alarm levels is satisfied, the corresponding alarm handling method of the alarm trigger mechanism can be executed to achieve intelligent response and handling of alarms. Optionally, the alarm handling method can include starting an additional redundant power supply, notifying relevant personnel, etc.

[0076] Through this embodiment, different energy strategies can be formulated according to different business types, ensuring business requirements while optimizing the overall energy efficiency; in addition, automatically executing the alarm handling method can intelligently handle faults occurring during work, improve the response speed and handling efficiency of abnormal events, and effectively prevent business interruptions.

[0077] The power supply control method of the server in the embodiment of the present application will be explained below with reference to optional examples. In this optional example, the above method is executed by the power supply control system of the server, and the power supply control system of the server can be used for 2U servers.

[0078] Figure 4 is a schematic structural diagram of the power supply control system of the server in this optional example, as Figure 4 shown, the power supply control system of the server can include six functional modules: a power consumption data acquisition module, a load prediction engine, a dynamic power distribution module, a redundant policy adaptive adjustment module, an energy efficiency optimization scheduler, and a central management control unit, as well as multiple power modules controlled by the system.

[0079] Optionally, the load prediction engine can output the load trend for the next 30 minutes every 5 minutes. The central management control unit can receive the prediction results, call the energy efficiency optimization scheduler to analyze the power supply strategy. The energy efficiency optimization scheduler can link with the redundant policy adaptive adjustment module to determine whether to perform PSU redundant degradation. After that, the central management control unit can send the results to the power control interface layer for actual power scheduling. In addition, during system operation, there can be a monitoring module continuously reporting power consumption and prediction errors back to the central management control unit, and the load prediction engine and the energy efficiency optimization scheduler can automatically adjust to optimize the control effect.

[0080] Optionally, the above dynamic power distribution module can also support hot plugging and automatic switching of PSUs; the above redundant policy adaptive adjustment module can also provide log auditing and policy rollback capabilities; the above central management control unit can be deployed on a BMC chip, running a lightweight Linux system, and support docking with the DataCenter Infrastructure Management (DCIM) / Energy Management System (EMS) of the data center, and can also provide a Web interface / Application Programming Interface (API) for remote policy configuration and alarm handling.

[0081] Taking the power supply control of a 2U dual-power server by this system as an example, 3 1.6kW PSUs can be configured (typically in an N+1 structure). When the predicted load drops below 20% at night, the system automatically shuts down 2 PSUs, and the main PSU supplies power independently; the standby PSU can be woken up 5 minutes in advance (when the trend is predicted) when the daytime traffic increases; after testing, the overall efficiency is increased from the original about 82% to about 91%, and the annual power saving is about 120 kWh per unit.

[0082] This system introduces a load prediction algorithm, a dynamic start-stop redundant power supply module, and an intelligent switching power distribution structure, and supports automatic switching of multiple redundant modes. Each module collaborates to complete the dynamic start-stop of the power supply module and the adjustment of the redundant policy, and can automatically balance the contradiction between redundancy and energy efficiency. The central management control unit can adopt BMC embedded control and support remote control through Redfish / IPMI interfaces. Integrated with the data center DCIM system / EMS platform, it can achieve compatibility between centralized / distributed control and improve the compatibility of the system.

[0083] Through this optional example, by predicting the load and dynamically starting and stopping redundant power supply modules, the PSU always operates in the optimal efficiency area, reducing the time that the PSU operates in the no-load or low-efficiency intervals, which can reduce power waste and improve the overall operating efficiency of the PSU; the system can automatically switch between the static redundant power supply mode and the dynamic redundant power supply mode according to the service SLA level, load trend, and alarm status, which can reduce the resource waste caused by excessive redundancy on the premise of ensuring safety; combined with the deep learning model, the load can be predicted with high precision, and the PSU can be woken up / slept in advance to avoid power supply jitter caused by delay; in addition, in terms of economic effects, power costs can be saved, and equipment aging and maintenance costs can be reduced; in terms of operation and maintenance and management benefits, the intelligent energy-saving strategy is transparent and controllable, and can also be seamlessly docked with the data center platform in related technologies, which can reduce the operation and maintenance pressure.

[0084] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0085] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as a read-only memory (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to enable a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of this application.

[0086] According to another aspect of the embodiments of this application, a power supply control device for a server is further provided. This power supply control device for a server can be used to implement the power supply control method for the server provided in the above embodiments, and those that have been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0087] Figure 5 is a structural block diagram of a power supply control device for an optional server according to an embodiment of the present application. As Figure 5 shown in the figure, the power supply control device of the server includes:

[0088] An acquisition unit 502, configured to acquire power consumption data of the set of power supplies within a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period ending at the current moment or a historical moment before the current moment;

[0089] A prediction unit 504, configured to predict the load trend of the set of power supplies within a specified future period based on the historical power consumption data to obtain the predicted load trend of the set of power supplies, where the specified future period is a future period starting at the current moment or a future moment after the current moment;

[0090] A switching unit 506, configured to dynamically switch the power supply mode of the set of power supplies according to the predicted load trend and specified mode switching conditions.

[0091] It should be noted that the acquisition unit 502 in this embodiment can be used to execute the above step S202, the prediction unit 504 in this embodiment can be used to execute the above step S204, and the switching unit 506 in this embodiment can be used to execute the above step S206.

[0092] Through the embodiment provided by the present application, the server has a set of power supplies. The power supply modes of the set of power supplies include a static redundant power supply mode and a dynamic redundant power supply mode. Among them, in the static redundant power supply mode, the number of redundant power supplies in the set of power supplies is fixed, and the redundant power supplies are in a standby state when not enabled; in the dynamic redundant power supply mode, the number of redundant power supplies in the set of power supplies is dynamically adjusted according to the load of the set of power supplies, and the redundant power supplies are in an off state when not started; the method includes: acquiring power consumption data of a set of power supplies within a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period ending at the current moment or a historical moment before the current moment; predicting the load trend of a set of power supplies within a specified future period based on the historical power consumption data to obtain the predicted load trend of a set of power supplies, where the specified future period is a future period starting at the current moment or a future moment after the current moment; dynamically switching the power supply mode of a set of power supplies according to the predicted load trend and specified mode switching conditions, solving the technical problems of low efficiency of power supplies and serious waste of electric energy existing in the power supply control method in the related art, improving the efficiency of power supplies, and reducing the waste of electric energy.

[0093] In an exemplary embodiment, the specified mode switching condition includes a first switching condition for switching from a static redundant power supply mode to a dynamic redundant power supply mode. The first switching condition includes that the duration for which the load of a set of power supplies is less than a first load threshold is greater than or equal to a specified time threshold. The switching unit includes: a first switching module, configured to switch the power supply mode of a set of power supplies to the dynamic redundant power supply mode when the power supply mode of the set of power supplies is the static redundant power supply mode and the predicted load trend indicates that the duration for which the predicted load of the set of power supplies is less than the first load threshold within a specified future period is greater than or equal to the specified time threshold.

[0094] In an exemplary embodiment, the specified mode switching condition includes a second switching condition for switching from the dynamic redundant power supply mode to the static redundant power supply mode. The second switching condition includes that the load of a set of power supplies is greater than or equal to a second load threshold, where the second load threshold is greater than the first load threshold. The switching unit includes: a second switching module, configured to switch the power supply mode of a set of power supplies to the static redundant power supply mode when the power supply mode of the set of power supplies is the static redundant power supply mode and the predicted load trend indicates that there is a future moment when the predicted load of a set of power supplies is greater than or equal to the second load threshold within a specified future period.

[0095] In an exemplary embodiment, the prediction unit includes: an extraction module, configured to perform feature extraction on historical power consumption data to obtain a historical power consumption sequence; an input module, configured to input the historical power consumption sequence, or the historical power consumption sequence and at least one of the following features into a load prediction model to obtain a predicted load trend output by the load prediction model: a historical power consumption load sequence of a set of power supplies within a specified historical period, a historical utilization rate sequence of the power consumption load of a server within a specified historical period, a task queue feature processed by the power consumption load of a server within a specified historical period, where the load prediction model is a time series model for performing load prediction.

[0096] In an exemplary embodiment, after predicting the load trend of a set of power suppliers within a specified future period based on historical power consumption data, the above-mentioned device further includes: a first startup unit, configured to, when the power supply mode of the set of power suppliers is a dynamic redundant power supply mode, start a specified main power supplier in the set of power suppliers when the predicted load trend of the set of power suppliers indicates that the predicted load of the set of power suppliers is less than or equal to a third load threshold, wherein other power suppliers in the set of power suppliers except the main power supplier are in a closed state as redundant power suppliers; a second startup unit, configured to, when the predicted load trend of the set of power suppliers indicates that the predicted load of the set of power suppliers gradually increases from being less than or equal to the third load threshold, start a number of redundant power suppliers corresponding to the load threshold in a set of load thresholds at a specified moment before the predicted load of the set of power suppliers reaches the load threshold in the set of load thresholds, wherein the time interval between the specified moment and the predicted moment when the predicted load of the set of power suppliers reaches the load threshold in the set of load thresholds is a specified time interval.

[0097] In an exemplary embodiment, the above-mentioned device further includes: a first determination unit, configured to determine a specified load ratio of the power suppliers in a set of power suppliers based on a specified power supply efficiency curve, wherein the specified power supply efficiency curve is used to indicate the energy conversion efficiency of the power suppliers in the set of power suppliers at the load ratio in a set of load ratios, and the specified load ratio corresponds to the highest energy conversion efficiency in the specified power supply efficiency curve; a control unit, configured to control the power suppliers in the set of power suppliers in a working state to supply power according to the specified load ratio.

[0098] In an exemplary embodiment, after inputting a historical power consumption sequence, or a historical power consumption sequence and at least one of the following features into a load prediction model, the above-mentioned device further includes: a calculation unit, configured to calculate a set of energy efficiency indicators of the server based on historical power consumption data, wherein the set of energy efficiency indicators includes at least one of the following: power utilization efficiency, cooling energy efficiency ratio; a generation unit, configured to generate a set of candidate operation modes based on the set of energy efficiency indicators of the server, wherein one of the candidate operation modes in the set of candidate operation modes is used to indicate a redundant configuration mode of a set of power suppliers, the number of power suppliers in the set of power suppliers in a working state, and the load distribution mode of the power suppliers in the set of power suppliers in a working state; a first execution unit, configured to determine a target operation mode from the set of candidate operation modes based on the predicted load trend, and perform operation control on the set of power suppliers according to the target operation mode.

[0099] An embodiment of the present application further provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any of the above-mentioned embodiments of the power supply control method for a server.

[0100] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps in any of the above-mentioned embodiments of the power supply control method for a server when running.

[0101] In an exemplary embodiment, the above-mentioned computer-readable storage medium may include, but is not limited to: various media such as USB flash drives, ROMs, RAMs, mobile hard disks, magnetic disks, or optical discs that can store computer programs.

[0102] An embodiment of the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above-mentioned embodiments of the power supply control method for a server.

[0103] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above-mentioned embodiments of the power supply control method for a server.

[0104] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0105] The above has introduced in detail a power supply control method, device, and electronic device provided by the present application. Specific examples are used herein to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A power supply control method for a server, characterized in that The server has a set of power supplies, and the power supply modes of the set of power supplies include a static redundancy power supply mode and a dynamic redundancy power supply mode. Among them, in the static redundancy power supply mode, the number of redundant power supplies in the set of power supplies is fixed, and when not enabled, the redundant power supplies are in a standby state; in the dynamic redundancy power supply mode, the number of the redundant power supplies in the set of power supplies is dynamically adjusted according to the load of the set of power supplies, and when not started, the redundant power supplies are in an off state; the method includes: Obtain the power consumption data of the set of power supplies in a specified historical period to obtain historical power consumption data, where the specified historical period is a historical period with the current moment or a historical moment before the current moment as the end moment; Based on the historical power consumption data, predict the load trend of the set of power supplies in a specified future period to obtain the predicted load trend of the set of power supplies, where the specified future period is a future period with the current moment or a future moment after the current moment as the start moment; Dynamically switch the power supply mode of the set of power supplies according to the predicted load trend and the specified mode switching conditions.

2. The method according to claim 1, characterized in that, The specified mode switching conditions include a first switching condition for switching from the static redundancy power supply mode to the dynamic redundancy power supply mode, and the first switching condition includes that the duration for which the load of the set of power supplies is less than the first load threshold is greater than or equal to the specified time threshold; The dynamically switching the power supply mode of the set of power supplies according to the predicted load trend and the specified mode switching conditions includes: When the power supply mode of the set of power supplies is the static redundancy power supply mode and the predicted load trend indicates that the duration for which the predicted load of the set of power supplies in the specified future period is less than the first load threshold is greater than or equal to the specified time threshold, switch the power supply mode of the set of power supplies to the dynamic redundancy power supply mode.

3. The method according to claim 2, wherein The specified mode switching conditions include a second switching condition for switching from the dynamic redundancy power supply mode to the static redundancy power supply mode, and the second switching condition includes that the load of the set of power supplies is greater than or equal to a second load threshold, where the second load threshold is greater than the first load threshold; The dynamically switching the power supply mode of the set of power supplies according to the predicted load trend and the specified mode switching conditions includes: When the power supply mode of the set of power supplies is the static redundancy power supply mode and the predicted load trend indicates that there is a future moment in the specified future period when the predicted load of the set of power supplies is greater than or equal to the second load threshold, switch the power supply mode of the set of power supplies to the static redundancy power supply mode.

4. The method according to claim 1, characterized in that Predicting the load trend of the set of power suppliers within a specified future period based on the historical power consumption data to obtain the predicted load trend of the set of power suppliers, including: Performing feature extraction on the historical power consumption data to obtain a historical power consumption sequence; Inputting the historical power consumption sequence, or the historical power consumption sequence and at least one of the following features into a load prediction model to obtain the predicted load trend output by the load prediction model: the historical power consumption load sequence of the set of power suppliers during the specified historical period, the historical utilization rate sequence of the power consumption load of the server during the specified historical period, the task queue characteristics processed by the power consumption load of the server during the specified historical period, where the load prediction model is a time series model for performing load prediction.

5. The method according to claim 1, wherein After predicting the load trend of the set of power suppliers within a specified future period based on the historical power consumption data, the method further includes: When the power supply mode of the set of power suppliers is the dynamic redundant power supply mode, when the predicted load trend of the set of power suppliers indicates that the predicted load of the set of power suppliers is less than or equal to a third load threshold, starting the specified main power supplier in the set of power suppliers, where the other power suppliers in the set of power suppliers except the main power supplier are in the off state as redundant power suppliers; When the predicted load trend of the set of power suppliers indicates that the predicted load of the set of power suppliers gradually increases from being less than or equal to the third load threshold, at a specified moment before the predicted load of the set of power suppliers reaches the load threshold in a set of load thresholds, starting the corresponding number of redundant power suppliers corresponding to the load threshold in the set of load thresholds, where the time interval between the specified moment and the predicted moment when the predicted load of the set of power suppliers reaches the load threshold in the set of load thresholds is a specified time interval.

6. The method according to claim 1, wherein The method further includes: Determining the specified load ratio of the power suppliers in the set of power suppliers based on a specified power supply efficiency curve, where the specified power supply efficiency curve is used to indicate the energy conversion efficiency of the power suppliers in the set of power suppliers at the load ratio in a set of load ratios, and the specified load ratio corresponds to the highest energy conversion efficiency in the specified power supply efficiency curve; Controlling the power suppliers in the set of power suppliers in the working state to supply power according to the specified load ratio.

7. The method according to claim 4, characterized in that, After inputting the historical power consumption sequence, or the historical power consumption sequence and at least one of the following features into the load prediction model, the method further includes: Calculating a set of energy efficiency indicators of the server based on the historical power consumption data, where the set of energy efficiency indicators includes at least one of the following: power utilization efficiency, cooling energy efficiency ratio; Generate a set of candidate operating modes based on the set of energy efficiency metrics of the server, wherein one candidate operating mode in the set of candidate operating modes is used to indicate the redundancy configuration method of the set of power supplies, the number of power supplies in the set of power supplies that are in the working state, and the load distribution method of the power supplies in the set of power supplies that are in the working state; Based on the predicted load trend, determine a target operating mode from the set of candidate operating modes, and perform operating control on the set of power supplies according to the target operating mode.

8. The method according to claim 1, wherein The method further includes: Based on the configuration information, determine a set of service level agreement rules and a set of alarm levels, wherein the set of service level agreement rules corresponds to multiple nodes powered by the set of power supplies, and indicates the minimum number of power supplies in the standby state, the minimum reserved power supply number, and the alarm threshold corresponding to the multiple nodes, and the set of alarm levels includes multiple alarm triggering mechanisms and the alarm handling methods corresponding to the multiple alarm triggering mechanisms; When one of the alarm triggering mechanisms in the set of alarm levels is satisfied, execute the alarm handling method corresponding to the alarm triggering mechanism.

9. A power supply control device for a server, characterized in that, The server has a set of power supplies, and the power supply modes of the set of power supplies include a static redundant power supply mode and a dynamic redundant power supply mode. In the static redundant power supply mode, the number of redundant power supplies in the set of power supplies is fixed, and when not enabled, the redundant power supplies are in the standby state; in the dynamic redundant power supply mode, the number of the redundant power supplies in the set of power supplies is dynamically adjusted according to the load of the set of power supplies, and when not started, the redundant power supplies are in the off state; the device includes: An acquisition unit, configured to acquire the power consumption data of the set of power supplies in a specified historical period to obtain historical power consumption data, wherein the specified historical period is a historical period ending at the current moment or a historical moment before the current moment; A prediction unit, configured to predict the load trend of the set of power supplies in a specified future period based on the historical power consumption data to obtain the predicted load trend of the set of power supplies, wherein the specified future period is a future period starting at the current moment or a future moment after the current moment; A switching unit, configured to dynamically switch the power supply mode of the set of power supplies according to the predicted load trend and the specified mode switching condition.

10. An electronic device, characterized in that, Includes: A memory, configured to store a computer program; A processor, configured to implement the steps of the power supply control method of the server according to any one of claims 1 to 8 when executing the computer program.

Images