Multi-node server power-on system, control method and multi-node server
By implementing distributed collaborative decision-making and dynamic power-on control in a multi-node server power-on system, the power surge problem during concentrated power-on of rack-mounted servers was solved, achieving orderly power-on and improved stability of the load.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-03-20
AI Technical Summary
The high-power surge generated when rack servers are powered on under concentrated load may damage other normally operating racks.
The system employs a multi-node server power-on mechanism. Through distributed collaborative decision-making and dynamic power-on control, each node makes collaborative decisions based on its own real-time power consumption data and the real-time power consumption data of its neighboring nodes, generating a power-on sequence that adapts to the global power consumption state, thus avoiding the power surge caused by the simultaneous startup of multiple node loads.
It enables orderly power-on of the load, improves energy efficiency and operational stability, avoids power surges, and protects the server power module and key components.
Smart Images

Figure CN121367313B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of server power supply, and in particular to a multi-node server power-on system, a control method and a multi-node server. BACKGROUND
[0002] With the explosive growth of artificial intelligence applications, the whole cabinet server is highly integrated with core components such as computing nodes, switching nodes, power management and heat dissipation systems in a standard cabinet for artificial intelligence computing scenarios to meet the stringent requirements of high-density computing power, low-latency communication and efficient heat dissipation.
[0003] In related technologies, the power supply module of the whole cabinet server adopts a power brick module parallel architecture, which follows the N+1 redundancy standard in design, but this method will cause a high power impact on the local power grid when the whole cabinet is powered on at the same time, and in severe cases, it may even damage other normally operating whole cabinets. SUMMARY
[0004] The present application provides a multi-node server power-on system, a control method and a multi-node server to at least solve the problem of high power impact when the load is concentrated in related technologies.
[0005] The present application provides a multi-node server power-on system, which comprises: a plurality of nodes, each node comprising a power-on controller, a communication interface, a voltage controller, a power consumption monitoring module and a timing logic indication module, wherein the power-on controller is connected with the communication interface, the voltage controller, the power consumption monitoring module and the timing logic indication module, the power consumption monitoring module is connected with the load, and the voltage controller is connected with the load, each node is connected with an adjacent node through the communication interface to form a connected network, wherein the power-on controller is configured to: in response to a power-on signal, acquire first power consumption data of the current node through the power consumption monitoring module and acquire second power consumption data of the adjacent node through the communication interface; determine whether the current node meets a preset power-on condition according to the first power consumption data and the second power consumption data; in response to the current node meeting the preset power-on condition, generate a power-on sequence of the current node according to the first power consumption data and the second power consumption data; and control the voltage controller according to the power-on sequence to power on the load.
[0006] The present application also provides a control method of a multi-node server power-on system, comprising: in response to a power-on signal, acquiring first power consumption data of the current node through the power consumption monitoring module and acquiring second power consumption data of the adjacent node through the communication interface; determining whether the current node meets a preset power-on condition according to the first power consumption data and the second power consumption data; in response to the current node meeting the preset power-on condition, generating a power-on sequence of the current node according to the first power consumption data and the second power consumption data; and controlling the voltage controller according to the power-on sequence to power on the load.
[0007] The application also provides a multi-node server comprising the multi-node server power-on system.
[0008] According to the application, since the power-on controllers of the nodes can make a cooperative decision based on the real-time power consumption data of the nodes and the real-time power consumption data of the adjacent nodes, dynamically generate a power-on sequence that adapts to the global power consumption state, avoid power superposition impact caused by simultaneous start of multiple node loads, realize orderly power-on of the loads, improve energy utilization efficiency and operation stability, and solve the problem of high power impact during centralized power-on of the loads in the related art. BRIEF DESCRIPTION OF DRAWINGS
[0009] In order to more clearly illustrate the embodiments of the application, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.
[0010] Figure 1 A connection diagram of a multi-node server power-on system according to an embodiment of the application;
[0011] Figure 2 A connection topology diagram of multiple nodes according to an embodiment of the application;
[0012] Figure 3 A work flow diagram of a multi-node server power-on system according to an embodiment of the application;
[0013] Figure 4 A block diagram of a multi-node server according to an embodiment of the application. DETAILED DESCRIPTION
[0014] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.
[0015] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. 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.
[0016] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0017] The multi-node server power-on system of this application embodiment will be described in detail below with reference to the accompanying drawings.
[0018] like Figure 1 As shown, the multi-node server power-on system 100 of this application embodiment includes: multiple nodes 10, each node including a power-on controller 11, a communication interface 12, a voltage controller 13, a power consumption monitoring module 14, and a timing logic indication module 15.
[0019] The power-on controller 11 is connected to the communication interface 12, the voltage controller 13, the power consumption monitoring module 14, and the timing logic indicator module 15, respectively. The power consumption monitoring module 14 is connected to the load, the voltage controller 13 is connected to the load, and each node is connected to the adjacent node through the communication interface 12 to form a connected network.
[0020] The power-on controller is configured to: in response to a power-on signal, acquire first power consumption data of the current node through a power consumption monitoring module and second power consumption data of adjacent nodes through a communication interface; determine whether the current node meets preset power-on conditions based on the first and second power consumption data; in response to the current node meeting the preset power-on conditions, generate a power-on sequence for the current node based on the first and second power consumption data; and control the voltage controller according to the power-on sequence to power on the load.
[0021] Specifically, each node is a functional unit with independent power-on control capability, and the node integrates a power-on controller, a communication interface, a voltage controller, a power consumption monitoring module and a timing logic indication module. The power-on controller is the decision unit of the node, which coordinates the operation of all modules, the communication interface undertakes the data interaction function between nodes, provides a channel for power consumption information sharing between adjacent nodes, the voltage controller is connected with the load and is responsible for executing the voltage output instruction in the power-on sequence, the power consumption monitoring module is connected with the load and collects the power consumption state of the load in real time to provide power consumption data for power-on decision, and the timing logic indication module can feedback the power-on progress and node state in real time, which is convenient for system debugging and operation and maintenance. Each node is connected with adjacent nodes through the communication interface to form a connected distributed collaborative network. Unlike the traditional centralized control architecture, each node can quickly obtain the running state of the surrounding nodes.
[0022] When the system receives the power-on signal, each node also receives the power-on signal. The power-on controller of the current node responds to the power-on signal and collects the first power consumption data of the current node through the power consumption monitoring module. At the same time, the power-on controller sends a power consumption query instruction to all adjacent nodes through the communication interface. After receiving the instruction, the adjacent nodes feed back their current power consumption state as the second power consumption data to the current node.
[0023] After obtaining the first power consumption data and the second power consumption data of all adjacent nodes, the power-on controller enters the decision stage and determines whether the current node meets the preset power-on condition according to the first power consumption data and the second power consumption data. The preset power-on condition can be set based on the principles of system overall power supply capacity, node-to-node power consumption threshold allocation, and avoidance of power superposition impact, for example, the sum of the total power consumption of the current node and all adjacent nodes is less than the system power supply peak threshold, or the number of nodes in the adjacent nodes that are executing power-on does not exceed the preset upper limit, or the sum of the startup peak value in the first power consumption data of the current node and the power margin occupied by the adjacent nodes does not exceed the safety threshold, etc.
[0024] The power-on controller aggregates and calculates the two types of collected data through an embedded algorithm to determine whether the preset condition is met. For example, the instantaneous peak value of the starting power consumption in the first power consumption data of the current node is 600 W, and the current actual power consumption of the three adjacent nodes in the second power consumption data is 200 W, 150 W, and 250 W, respectively. The aggregated calculation of the two types of data is 600 W + 200 W + 150 W + 250 W = 1.2 KW. However, the maximum instantaneous power supported by the system as a whole is only 1 KW. Obviously, this case does not meet the above-mentioned preset condition. On the contrary, if the total power consumption calculated by aggregation is less than the maximum instantaneous power supported by the whole system, it meets the preset power-on condition. It should be noted that the current actual power consumption of one of the adjacent nodes in the second power consumption data is 250 W. This value can be the actual power consumption of the node itself or the sum of the actual power consumption of the adjacent nodes.
[0025] In response to the current node meeting the preset power-on condition, the current node starts to power on, and according to the load characteristics of the first power consumption data and the adjacent node power consumption redundancy reflected by the second power consumption data, the power-on sequence of the current node is dynamically generated. The power-on sequence is the process in which each power supply branch is gradually powered on in a preset order, voltage amplitude, and time interval during the power-on startup process of the node. The internal power-on controller, voltage controller, power consumption monitoring module, and timing logic indication module of the node need to be powered on in sequence according to the power-on sequence, rather than simultaneously, thereby reducing the power impact.
[0026] The voltage controller is controlled according to the power-on sequence. After receiving the instruction of the power-on sequence, the voltage controller can control the amplitude and rising rate of the output voltage through pulse width modulation or other methods to provide smooth power-on power supply for the load. At the same time, the power consumption monitoring module can continuously collect the actual power consumption data of the load after power-on. If an abnormality is found, it is immediately fed back to the power-on controller. The controller can pause the power-on process or adjust the voltage output to ensure safe operation.
[0027] The power-on sequence is not a fixed timing template, but a targeted voltage output control scheme. For example, if the current power consumption of the adjacent node is low and the system power supply margin is sufficient, the power-on sequence can adopt a mode of fast voltage boost and steady-state maintenance. The voltage controller quickly raises the output voltage to the rated value of the load according to the sequence instruction, shortens the power-on time, and if the adjacent node has occupied part of the power resource, the power-on sequence can adopt a step-by-step voltage boost strategy, which divides the voltage into multiple gradients and gradually raises it. Each gradient is maintained for a preset time to slowly increase the load startup power consumption and avoid the power consumption of the adjacent node to form an impact. For loads with high startup power consumption peak value, the power-on sequence can also add a pre-charge phase, which first pre-charges the load capacitor through low voltage, and then gradually raises it to the rated voltage, further reducing the startup power impact.
[0028] During the whole power-on process, the timing logic indication module synchronously feeds back the node state, so as to facilitate the operator to master the system operation in real time.
[0029] Suppose that a server internally contains n nodes, when the power-on controller of one of the nodes judges that the preset power-on condition is met, the current node starts to power on in sequence according to the power-on sequence, and then, the adjacent node connected to the node calculates whether the total power consumption after joining meets the preset power-on condition according to the power consumption change data transmitted, if yes, starts to power on itself, and sequentially promotes to the next adjacent node connected, and performs power-on control step by step.
[0030] Each node dynamically judges whether the system currently has sufficient power redundancy to perform further power-on operation according to the power collection information from the adjacent node and the power demand of the node itself, and the chain-like cooperative control mechanism not only effectively avoids the problem of instantaneous power overload, but also enhances the stability and intelligent degree of the system.
[0031] Therefore, the multi-node server power-on system of the embodiment realizes the orderly start of all loads, avoids the power impact caused by centralized power-on, effectively protects the key components such as the server power module and the power bus, reduces unnecessary peak power reserve, and improves the energy utilization efficiency and the overall operation stability of the system.
[0032] In some embodiments of the present application, the communication interfaces are connected through a communication bus, and the communication interfaces include multiple communication buses, wherein the current node establishes a connection with multiple different adjacent nodes through the communication interfaces.
[0033] Specifically, a distributed interconnected network is constructed between the nodes through the communication interfaces and the communication bus, the communication interfaces are connected through the communication bus, multiple independent communication buses are integrated, and hardware support is provided for parallel communication of the current node and the adjacent nodes. The type of the communication bus can be an I2C bus, the I2C bus supports one master and multiple slaves or multiple masters and multiple slaves topology structure, the current node can establish a communication connection with multiple different adjacent nodes through the same bus, and each node has a unique address for identity recognition. For example, a communication interface includes three I2C buses, and at most three different adjacent nodes can be connected.
[0034] In the data transmission, the power-on controller of the current node sends the power consumption query instruction to the target adjacent node through the specified bus channel, and the adjacent node feeds back its own power consumption data after receiving through the corresponding bus channel. The parallel transmission of the multiple buses enables the current node to synchronously and in real time acquire the power consumption data of all adjacent nodes, which is used for collaborative decision of the power-on sequence of the entire system, greatly shortens the communication delay, and improves the system communication reliability.
[0035] The embodiment provides parallel communication hardware support for the current node and multiple adjacent nodes to build a distributed interconnected network by integrating multiple independent communication buses in the communication interface. In the data transmission, the current node can synchronously send the power consumption query instruction to each adjacent node through the multiple buses and receive the feedback, which greatly shortens the communication delay, synchronously and in real time acquires the power consumption data of all adjacent nodes to efficiently and collaboratively decide the power-on sequence of the system, and avoids the problem of overall communication interruption caused by single bus failure, thereby significantly improving the system communication reliability in combination with the unique address identification of the nodes.
[0036] In some embodiments of the present application, the multiple nodes form an undirected connected network through the communication interface.
[0037] Specifically, Figure 2 As shown, the connection topology between the multiple nodes is shown, and 12 nodes are taken as an example to illustrate the connection topology relationship between the multiple nodes. As shown in the figure, the nodes include N1, N2, …, N12, the adjacent nodes of the N1 node are N2, N5, and N9, the adjacent nodes of the N2 node are N1 and N3, and so on. The adjacent nodes are connected through the communication interface, and the communication process is bidirectional. The N1 node can acquire the power consumption data of the N2 node, the N2 node can also acquire the power consumption data of the N1 node, and the N2 node can also acquire the power consumption data of the N3 node. Therefore, with the connectivity, the N1 node can indirectly acquire the power consumption data of the N3 node. Any node in the topology relationship can access all other nodes and acquire the power consumption information of other nodes in the entire system. Thus, the network formed is an undirected connected network. The undirected connected characteristic enables each node to equally acquire the global power consumption information, avoids the single-point failure of centralized control, and ensures that the power-on sequence dynamically adapts to the entire network state through distributed data sharing and collaborative decision.
[0038] The multiple nodes of the embodiment form an undirected connected network through the communication interface, which enables each node to equally acquire the global power consumption information, avoids the single-point failure risk of centralized control, and ensures that the power-on sequence dynamically adapts to the entire network state through distributed data sharing and collaborative decision. In addition, the bidirectional communication between the adjacent nodes also guarantees the reliability of the power consumption data interaction.
[0039] In some embodiments of the present application, the power-on controller is further configured to: in response to the current node not satisfying the preset power-on condition, continuously acquire second power consumption data of adjacent nodes through the communication interface; and continue to determine whether the current node satisfies the preset power-on condition according to the first power consumption data and the continuously acquired second power consumption data until the current node satisfies the preset power-on condition.
[0040] Specifically, when the power-on controller determines that the current node does not satisfy the preset power-on condition, a dynamic waiting and retry process is triggered, the current node enters a waiting state, and the power-on controller continuously sends a power consumption state query instruction to all adjacent nodes through the communication interface every preset time (e.g., 100 ms) to update the second power consumption data in real time, and synchronously refreshes the first power consumption data through the power consumption monitoring module. According to the first power consumption data and the continuously acquired second power consumption data, it is determined again whether the current node satisfies the preset power-on condition, and the determination logic is consistent with the foregoing, until the current node satisfies the preset power-on condition and is powered on.
[0041] This embodiment improves the dynamic accuracy of power-on condition determination by continuously acquiring second power consumption data of adjacent nodes and synchronously refreshing first power consumption data of the node itself. Through the dynamic waiting and retry process, the overall power consumption balance of the system can be ensured, the risk of overload can be reduced, and the system operation stability can be improved.
[0042] In some embodiments of the present application, the second power consumption data includes power consumption data of all relevant nodes except the current node in the plurality of nodes.
[0043] Specifically, the second power consumption data is not limited to only the direct adjacent nodes of the current node, but covers all relevant nodes such as the adjacent nodes of the current node and the associated nodes of the adjacent nodes. Figure 2 For example, when the N1 node acquires the second power consumption data, the second power consumption data includes not only the power consumption data of the N2, N5 and N9 nodes, but also the power consumption data of the N3, N4, N6, N7, N8, N10, N11 and N12 nodes, thereby ensuring the comprehensiveness of data acquisition.
[0044] When the power-on controller of the current node sends a power consumption query instruction to the direct adjacent nodes through the communication interface, it synchronously requests the adjacent nodes to forward the query instruction to their own adjacent nodes. After the adjacent nodes receive the instruction, on the one hand, they feed back their real-time power consumption data, and on the other hand, they pass the query instruction to the associated nodes. The associated nodes then return their power consumption data to the adjacent nodes, which are aggregated and fed back to the current node, so that the current node can comprehensively perceive the power consumption load of all relevant nodes except the current node.
[0045] The embodiment extends the data coverage range to the associated nodes of the adjacent nodes, enables the current node to master a more complete surrounding power consumption situation, realizes the comprehensiveness and integrity of power consumption data collection, reduces the communication link redundancy by transferring the power consumption data through the adjacent nodes, ensures the real-time of the data, enables the current node to more accurately perceive the surrounding power consumption load, and enables the current node to master the power consumption load of all related nodes, formulates a power-on strategy based on the global power consumption situation, and improves the energy efficiency ratio and operation stability of the entire system.
[0046] In some embodiments of the present application, the power-on controller is specifically configured to: analyze the first power consumption data and the second power consumption data respectively to obtain the current power consumption of each node in the plurality of nodes; calculate the power-on condition power consumption of the current node according to the current power consumption of each node; if the power-on condition power consumption is less than or equal to the maximum instantaneous power allowed by the multi-node server, it is determined that the current node satisfies the preset power-on condition; if the power-on condition power consumption is greater than the maximum instantaneous power, it is determined that the current node does not satisfy the preset power-on condition.
[0047] Specifically, the power-on controller analyzes the collected first power consumption data and second power consumption data. For the first power consumption data, the power-on instantaneous power consumption value of the current node load is extracted after analysis. For the second power consumption data, the current actual power consumption value of each related node is extracted one by one according to the node ID (Unique Identifier, Unique Identifier) classification analysis. It should be noted that the power-on instantaneous power consumption value described in this embodiment can be pre-stored in the storage space of the power consumption monitoring module. When the node corresponding to the power consumption monitoring module receives the power-on signal, it can be directly obtained from the storage space of the power consumption monitoring module.
[0048] Based on the analyzed full-quantity node power consumption data, the power-on condition power consumption of the current node is calculated, which is the sum of the instantaneous power consumption peak value required by the current node to start and the current actual power consumption value of all related nodes. For example, a server contains 8 computing nodes and 2 switching nodes. Assuming that the power consumption peak value of each computing node is 4000W and the power peak value of each switching node is 2000W, the current node is one of the computing nodes, and all nodes are in the power-on process when the current node obtains the second power consumption data. The power-on condition power consumption is: 8x4000W+2x2000W=36000W (i.e. 36KW).
[0049] The calculated power consumption under the power-on condition is compared with the maximum instantaneous power allowed by the multi-node server, wherein the maximum instantaneous power is a safety threshold preset based on parameters such as rated power of the power supply module and bus carrying capacity. If the power consumption under the power-on condition is less than or equal to the maximum instantaneous power, for example, the maximum instantaneous power supported by the system as a whole is 40 KW, and 36 KW is less than 40 KW, it is determined that the current node meets the preset power-on condition, and the power-on sequence generation process can be started. If the power consumption under the power-on condition is greater than the maximum instantaneous power, for example, the maximum instantaneous power supported by the system as a whole is 28 KW, and 36 KW is greater than 28 KW, it is determined that the current node does not meet the preset power-on condition, and the dynamic waiting and retry mechanism is triggered by the power-on controller, and the data is continuously updated and recalculated until the requirement is met.
[0050] This embodiment determines whether the current node meets the preset power-on condition by comparing the sum of the instantaneous peak value of the current node and the actual power consumption of all related nodes with the maximum instantaneous power, which can avoid power overload caused by total power consumption exceeding the threshold when a single node is powered on, maintain the stability of multi-node operation, ensure the safety of the power-on system, dynamically adapt to power consumption demand, and improve the efficiency of server resource utilization.
[0051] Further, the power-on controller is further configured to: obtain type information of the current node before obtaining the first power consumption data of the current node through the power consumption monitoring module; and perform subsequent operations after a preset time when it is determined according to the type information that the current node is a switching node.
[0052] Specifically, the subsequent operation refers to obtaining the first power consumption data of the current node through the power consumption monitoring module, before which the power-on controller obtains the type information of the current node through a hardware identification reading interface or a system configuration file. The type information is pre-fixed in the EEPROM (Electrically Erasable Programmable Read-Only Memory) or BIOS (Basic Input / Output System) of the node, and includes classification identifiers such as switching nodes and computing nodes.
[0053] The power-on controller analyzes and judges the read type information. If the current node is identified as a switching node, a delay mechanism is triggered, a built-in timer module is called, a preset delay time (such as 1s) is loaded, and a waiting state is entered. After the preset time countdown ends, the power-on controller starts the subsequent operation, collects the first power consumption data of the current node through the power consumption monitoring module, and if the current node is identified as a non-switching node (such as a computing node), no delay is needed, and the first power consumption data collection process is directly executed, thereby prioritizing the power-on of the computing node.
[0054] In addition, the preset time can be set by a person skilled in the art according to actual conditions, and is not specifically limited here.
[0055] The embodiment ensures that the computing node enters the available state quickly by differentiating the process priorities, and the exchange node collects the first power consumption data after the preset time, and the computing node does not need to delay, thereby ensuring the power-on and starting efficiency of the computing node, and meeting the requirement of the system for preferentially guaranteeing the core computing power supply.
[0056] In some embodiments of the present application, the timing logic indication module includes a plurality of light groups, and the timing logic indication module is configured to feed back the power-on process of the current node through the lighting combination of the plurality of light groups; and the power-on controller is further configured to control the plurality of light groups according to the power-on sequence.
[0057] Specifically, the timing logic indication module includes a plurality of light groups, and the timing logic indication module feeds back the power-on process of the current node through the lighting combination of the plurality of light groups. For example, the timing logic indication module is integrated on the input / output board card of each node, and 8 LED (Light Emitting Diode) light groups are used to form an 8-bit (1 byte) display unit, which accurately feeds back the power-on process of the current node and the system timing state through the differentiated lighting combination, and the control logic thereof is dynamically scheduled by the power-on controller according to the power-on sequence.
[0058] Taking 8 LED light groups as an example, the 8 LED light groups of the timing logic indication module are divided into two groups of low 4 bits (0-3 bits) and high 4 bits (4-7 bits) according to functions, the low 4-bit LED corresponds to the division of 4 functional groups (12 nodes are evenly distributed to 4 groups), which not only identifies the group to which the current node belongs, but also feeds back the power-on process and abnormal state in the group. When the power-on controller executes the power-on sequence, if the current node meets the preset power-on condition to enter the power-on phase according to the power-on sequence, the LED light corresponding to the group to which the current node belongs is controlled to be lit in green.
[0059] Meanwhile, the power-on controller collects the load power-on data in real time through the power consumption monitoring module, and if it detects abnormal voltage fluctuation, power consumption exceeding the threshold, or failure to complete the stage switching according to the power-on sequence instruction, it determines that the power-on is abnormal, and immediately switches the corresponding low 4-bit LED to red alarm. The operation and maintenance personnel can directly locate the fault node and the group to which it belongs through the LED color.
[0060] During troubleshooting, the red alarm identification of the low 4-bit LED can quickly lock the abnormal node, and through the log feedback of the power-on controller, it can further locate whether the problem is the load failure, the abnormal execution of the voltage controller, or the coordination problem of the adjacent nodes, thereby greatly shortening the troubleshooting time.
[0061] The high 4-bit LED is used to represent the complete power-on timing state code, and the 4-bit LED can combine 16 basic states, each state maps 2 power-on sequence core stages, covering 32 key timing nodes from pressing the start button to the system fully starting. When the power-on controller executes each stage of the power-on sequence, it will synchronously output the corresponding state code signal to drive the high 4-bit LED to light up in the preset combination (for example, a stage corresponds to the "0101" combination, and the 4th and 6th LEDs light up), which maps the current power-on timing node in real time.
[0062] The cooperative display design of the 8-bit LED not only realizes accurate identification of node grouping, process and exception through the low 4-bit, but also covers all timing nodes through the high 4-bit, so that the power-on process and state are visualized and traceable. The power-on controller is linked with the power-on sequence to ensure that the LED display is synchronized with the actual power-on state in real time, which provides intuitive support for operation and maintenance debugging and system monitoring.
[0063] The timing logic indication module of the embodiment includes a plurality of lamp groups, and the timing logic indication module feeds back the power-on process of the current node through the lighting combination of the plurality of lamp groups. The power-on controller is also used to control the plurality of lamp groups according to the power-on sequence, so as to realize the visualization of the power-on process and state, realize accurate monitoring of the key timing node state, quickly lock the fault range, shorten the exception troubleshooting period, reduce the operation complexity, and guarantee the reliability of the power-on system.
[0064] In some embodiments of the present application, the power-on controller is also used to receive feedback information of the voltage controller and control the plurality of lamp groups according to the feedback information and the power-on sequence.
[0065] Specifically, the voltage controller collects the output voltage amplitude, voltage rise rate, load access state and other key parameters as feedback information after performing each voltage output operation of the power-on sequence, and uploads the feedback information to the power-on controller through the communication bus. The feedback period is consistent with the stage switching period of the power-on sequence.
[0066] After receiving the feedback information, the power-on controller compares the feedback information with preset parameters of the current stage of the power-on sequence. If the feedback information shows that the voltage controller has accurately executed the instruction and there is no abnormal alarm, the controller controls the light group to update the state according to the preset logic. For example, when the power-on sequence advances to the processor power supply stage and the voltage controller feedbacks that the stage is executed, the power-on controller lights up the green LED light group corresponding to the node group in the low 4 bits, and simultaneously switches the high 4 bits LED to the time sequence state code combination corresponding to the stage. If the feedback information shows that the voltage output error exceeds the allowed range, the voltage rising rate is abnormal, or the load is not normally connected, it is determined that the power-on is abnormal, the controller immediately switches the LED corresponding to the low 4 bits group to red alarm, and simultaneously keeps the high 4 bits LED unchanged, so as to facilitate positioning of the specific power-on stage where the abnormality occurs.
[0067] In addition, when the voltage controller feedbacks the fault recovery information, the power-on controller recompares the feedback information with the power-on sequence parameters, and switches the red alarm LED back to the green display corresponding to the stage or updates the light group combination to the next stage according to the current power-on progress, to realize state synchronization after abnormal recovery.
[0068] The embodiment realizes closed-loop synchronization of the light group display and the actual power-on state by receiving real-time feedback information of the voltage controller by the power-on controller and accurately comparing the power-on sequence, which not only ensures visual and accurate presentation of the node power-on progress, time sequence stage and grouping information, but also quickly locates voltage output abnormality, stage execution failure and other problems, supports real-time updating of the state after abnormal recovery, effectively improves the monitorability, fault locating efficiency and system maintainability of the multi-node server power-on process, and guarantees the accuracy and reliability of the power-on sequence execution.
[0069] In some embodiments of the present application, the power-on controller is connected with the load. The power-on controller is further configured to receive power-on results corresponding to the power-on sequence fed back by the load, and optimize the power-on sequence according to the power-on results.
[0070] Specifically, the power-on controller is connected with the load. The power-on controller synchronously receives power-on result data fed back by the load in the process of executing the power-on sequence by the voltage controller, which includes actual power consumption curve, voltage tolerance range, start success or failure, and whether the steady-state operation meets the standard and other key information.
[0071] The power-on controller compares and analyzes the feedback power-on result with the target power-on sequence, thereby optimizing the power-on sequence according to the power-on result. For example, if the load feedback startup is successful but the peak power consumption is higher than expected, it indicates that the current power-on sequence has too fast a boost rate, and the controller will optimize the sequence parameters, lengthen the stage interval of the step-up boost or reduce the initial boost gradient, to avoid the load from bearing too large a power consumption impact during the next power-on, or if the load feedback voltage fluctuation amplitude exceeds the allowed range after the voltage stabilizes, the pulse width modulation parameters of the voltage controller are optimized, a voltage stabilization adjustment stage is added to the sequence, or if the load feedback startup fails or the stage switching times out, the power-on controller will combine the load fault code to check whether the voltage threshold setting of the power-on sequence is unreasonable, and then adjust the target voltage value of the corresponding stage or increase the pre-charge duration.
[0072] The optimized power-on sequence is stored in the local cache of the power-on controller as the default sequence for the next power-on of the load. For nodes supporting multi-load adaptation, the controller also establishes exclusive sequence archives for different loads to achieve one load-one optimized sequence.
[0073] This embodiment directly receives the actual power-on result of the load feedback through the power-on controller and dynamically optimizes the power-on sequence parameters based on the result, thereby achieving adaptive matching of the power-on sequence and the load characteristics, effectively solving the problem that the fixed power-on sequence is difficult to adapt to different load types or load aging parameters, improving the load power-on success rate and startup stability, reducing power consumption impact, startup failure and other abnormalities caused by sequence mismatch, and further improving energy utilization efficiency and long-term operation reliability and adaptability of the power-on system by continuously optimizing the power-on process to better meet the actual operation requirements of the load.
[0074] In some embodiments of the present application, the second power consumption data includes address information, power consumption information, power-on progress information, and next-step operation instruction information.
[0075] Specifically, the address information is used to identify the data source node, so that the receiving node can distinguish the specific node that sends the information. The power consumption information contains real-time power consumption data of the current node, which is used by the target node for power calculation and judgment. The power-on progress information shows the current power supply timing stage of the sending node, which facilitates global timing consistency control. The next-step operation instruction information represents the action to be performed by the sending node next time, providing a decision reference for other nodes.
[0076] The power-on controller of each node analyzes the power consumption change and the overall load of the system after receiving the second power consumption data, and determines the next strategy. For example, whether the current node should send its own voltage controller enable signal in advance to enter the next stage, or delay the operation to avoid system instability caused by excessive instantaneous power. When making the judgment, the power-on controller will compare the power-on progress of other nodes and the operation instructions to be executed, combine the local instantaneous power consumption status, and generate adaptive control commands to realize dynamic collaborative power-on at the system level, ensuring smooth transition of the server from standby state to working state.
[0077] This embodiment integrates address information, power consumption information, power-on progress information, and next operation instruction information through the second power consumption data, so that each node power-on controller can comprehensively and accurately perceive the identity of other nodes, real-time power consumption load, timing advancement state, and subsequent action planning, and realize system-level dynamic collaborative decision-making in combination with local status, avoiding power-on conflicts and instantaneous power overload between nodes caused by information asymmetry, ensuring smooth transition of the server from standby to working state. Through global timing consistency control and action prediction optimization, the orderliness, synchronicity, and system running stability of multi-node power-on are improved, the power impact risk is reduced, and the energy utilization efficiency is improved.
[0078] In some embodiments of the present application, the power-on controller is also used to optimize the power-on sequence according to the power-on progress information and the next operation instruction information.
[0079] Specifically, after the power-on controller receives the second power consumption data of each related node, it analyzes the power-on progress information and the next operation instruction information, and establishes a global timing atlas in combination with the current execution stage of the local power-on sequence, to optimize the power-on sequence. For example, if it is found that the power-on progress of the adjacent node lags behind and the next operation instruction is continuous voltage rise, the controller will optimize the local power-on sequence, such as prolonging the holding time of the current voltage gradient, reducing the voltage rise rate of the next stage, or temporarily delaying the start time of the high-power consumption stage, to avoid the superposition of power demand with adjacent nodes leading to system total power consumption exceeding the limit.
[0080] For scenarios that require timing coordination, the power-on controller will adjust the stage switching time of the local sequence according to the power-on progress information and the next operation instruction of most nodes to ensure global timing consistency. The optimized power-on sequence will be updated and executed in real time, and the local updated power-on progress and next operation instruction will be fed back to other nodes to ensure the orderly coordination of system-level power-on actions.
[0081] The power-on controller of this embodiment dynamically optimizes the local power-on sequence based on the power-on progress information and next operation instruction information of other nodes, realizes precise cooperation and global timing adaptation of power-on actions among multiple nodes, avoids power superposition impact and system instability problems caused by unsynchronized timing and operation conflicts among nodes, flexibly adjusts the power-on rate, stage interval and action timing by adapting the state of other nodes, improves the orderliness and synchronization of multi-node power-on, shortens the overall power-on period and reduces invalid power consumption through global cooperative optimization, ensures smooth transition from standby to working state of the system, and enhances the cooperation, stability and operation efficiency of the multi-node server power-on system.
[0082] In some embodiments of the present application, the multi-node server power-on system further comprises a temperature sensor connected to the power-on controller, the temperature sensor being arranged on the load for acquiring the temperature of the load and sending the temperature to the power-on controller.
[0083] Specifically, the multi-node server power-on system adds a temperature sensor as an important component for load state monitoring, which is signal-connected with the power-on controller and deployed on the load, and real-time collects temperature data of the load during the entire power-on process and sends it to the power-on controller. The temperature data provides a temperature dimension reference for the load running state for the power-on controller, so that the power-on controller can make comprehensive decisions in combination with the load temperature state when executing the power-on sequence, judging the power-on condition or optimizing the control strategy.
[0084] This embodiment realizes real-time collection and feedback of the load temperature by arranging a temperature sensor on the load and connecting it with the power-on controller, provides a temperature dimension state reference for the power-on controller, so that the controller can make comprehensive judgments in combination with temperature, power consumption, timing and other multi-dimensional information when making decisions, prevents the load from being damaged or running abnormally due to overheating during the power-on process, perfects the closed-loop control mechanism of the power-on system, and improves the safety, stability and reliability of the multi-node server power-on process.
[0085] In some embodiments of the present application, the power-on controller is further configured to: in response to the temperature being greater than a preset temperature threshold, control the current node to stop power-on through the voltage controller, and generate corresponding alarm information.
[0086] Specifically, the power-on controller stores the preset temperature threshold corresponding to each power-on stage of the load, and after receiving the load temperature data continuously transmitted by the temperature sensor, it compares it with the preset threshold of the current stage in real time. When it is detected that the load temperature exceeds the corresponding threshold, it is determined to be an overheating risk state, and a stop power-on instruction is sent to the voltage controller to terminate the execution of the power-on sequence of the current node by cutting off the voltage output or closing the power switch, preventing the load core components from being burned out due to continuous temperature rise.
[0087] Meanwhile, the power-on controller generates alarm information containing key information such as fault type, fault node ID, over-limit temperature value, and occurrence stage, which can be transmitted to the management platform through the communication bus or synchronized to the local timing logic indication module, so as to facilitate the operation and maintenance personnel to quickly locate the overheating node and fault scene and timely carry out cooling or maintenance operation. After the load temperature naturally falls below the safety threshold, the controller can trigger an automatic retry power-on process or wait for a manual reset instruction to restart the power-on sequence, ensuring the safety and controllability of the operation.
[0088] In addition, the preset temperature threshold can be set by a person skilled in the art according to the actual situation, which is not specifically limited here.
[0089] The power-on controller of this embodiment stops the current node power-on in time through the voltage controller and generates alarm information when detecting that the load temperature exceeds the preset threshold, realizing rapid response and active protection of the overheating risk, avoiding hardware damage or performance degradation of the load due to continuous overheating, and accurately feeding back the fault scene through the alarm information, which facilitates the operation and maintenance personnel to quickly locate and handle the problem, ensures the hardware safety and operation stability of the multi-node server, and improves the fault perceptibility and maintainability of the system.
[0090] In some embodiments of the present application, the multi-node server power-on system further comprises an alarm module connected with the power-on controller, for receiving the alarm information fed back by the power-on controller and providing the alarm information to the target user.
[0091] Specifically, the alarm module can integrate an audible and visual alarm unit and an information transmission interface, establish a connection with the power-on controller of each node, receive and analyze the alarm information fed back by the power-on controller, trigger audible and visual alarms, for example, issue intermittent alarm sounds through a buzzer and control the warning light to flash, realizing rapid early warning. The alarm information can also be uploaded to the server management platform or the mobile terminal of the target user, such as in the form of a short message, a push notification, etc., to synchronize the early warning, ensuring that the target user can timely know the fault condition whether or not he is on site, wherein the target user can be a management personnel. In addition, the alarm module also supports local storage of the latest alarm records, and the target user can query fault details through the small display screen or connected host computer of the module.
[0092] This embodiment realizes real-time reception and accurate push of alarm information by setting an alarm module connected with the power-on controller, can timely deliver fault information to the target user, shortens the fault response and troubleshooting time through rapid early warning, and improves the perceptibility of system faults and the convenience of operation and maintenance.
[0093] In some embodiments of the present application, the power-on controller is further configured to: in the process of controlling the voltage controller to power on the load according to the power-on sequence, acquire the current power of the load in real time through the power consumption monitoring module; and in response to the current power being greater than a preset power threshold, control the current node to stop powering on through the voltage controller.
[0094] Specifically, the power-on controller synchronously triggers the power consumption monitoring module to perform real-time data acquisition in the process of executing the power-on sequence and powering on the load through the voltage controller, acquires the current power of the load, compares the current power with a power threshold corresponding to each power-on stage of the preset load, determines a power overload state when the current power is greater than the preset power threshold, and sends an emergency stop instruction to the voltage controller. The voltage controller terminates the power-on process of the current node by cutting off the power supply loop, closing the power switch, or reducing the output voltage, thereby avoiding the burning of core components of the load or the damage of the circuit caused by an overload current.
[0095] In addition, the preset power threshold can be set by a person skilled in the art according to actual conditions, and is not specifically limited herein.
[0096] The power-on controller of this embodiment acquires the current power in real time through the power consumption monitoring module in the process of powering on the load, stops the current node from powering on in time through the voltage controller when it is detected that the power exceeds the preset threshold, realizes the rapid response and active protection of power overload, perfects the power safety management and control mechanism of the power-on system, and improves the stability and reliability of the multi-node server power-on process.
[0097] In some embodiments of the present application, the multi-node server power-on system further comprises a heat dissipation module connected with the power-on controller, wherein the power-on controller is further configured to control the heat dissipation module according to the power-on sequence.
[0098] Specifically, the heat dissipation module can adopt a temperature-controlled fan and be communicatively connected with the power-on controller through a pulse width modulation interface or a bus control mode, and can adjust the heat dissipation power according to the instruction of the power-on controller. The power-on controller pre-stores heat dissipation strategies matched with each stage of the power-on sequence, controls the heat dissipation module according to the power-on sequence, for example, controls the heat dissipation module to operate at low power before starting power-on, dynamically adjusts the working state of the heat dissipation module according to the power consumption characteristics and temperature feedback of the current power-on stage during the execution of the power-on sequence, and controls the heat dissipation module to switch to a constant temperature mode after the load completes power-on and enters a steady state, to adaptively adjust the heat dissipation intensity according to the real-time temperature and power consumption of the load, so as to meet the heat dissipation demand while reducing energy consumption.
[0099] The embodiment realizes the cooperative adaptation of the heat dissipation strategy and the power-on process, self-adapts the heat dissipation intensity, balances the heat dissipation effect and energy efficiency, and improves the overall energy efficiency of the power-on process of the multi-node server by connecting the heat dissipation module with the power-on controller, dynamically controlling the working state of the heat dissipation module according to the power-on sequence by the power-on controller.
[0100] In some embodiments of the present application, the power-on controller is further configured to: acquire real-time power consumption data of the current node through the power consumption monitoring module, and calculate a hazard degree value of the power generation of the current node to the multi-node server according to the real-time power consumption data; and if the hazard degree value is greater than a preset hazard threshold, control the current node to stop power-on through the voltage controller.
[0101] Specifically, the power-on controller acquires real-time power consumption data of the current node through the power consumption monitoring module during the power-on process of the current node, calculates a hazard degree value corresponding to the power generation of the current node through the built-in hazard degree quantification model, and compares the hazard degree value with a preset hazard threshold. If the hazard degree value is greater than the preset hazard threshold, it indicates that the real-time power consumption of the current node has posed a significant threat to the normal power-on of other nodes. The power-on controller immediately sends an emergency stop instruction to the voltage controller, and the voltage controller terminates the power-on process of the current node by cutting off the power supply loop, closing the power switch or reducing the output voltage, thereby avoiding the spread of hazards.
[0102] The embodiment acquires real-time power consumption data of the current node through the power consumption monitoring module and quantitatively calculates a hazard degree value of the current node to the multi-node server. When the hazard degree is greater than a preset threshold, the power-on of the current node is stopped in time through the voltage controller, which can accurately identify and block potential threats posed by the current node to the normal operation of other nodes, thereby improving the accuracy, reliability and overall stability of the safety protection of the power-on process of the multi-node server.
[0103] In some embodiments of the present application, the hazard degree value of the power generation of the current node to the multi-node server is calculated by the following formula:
[0104] C=ax 2 +bx+c;
[0105] Wherein, C is the hazard degree of the power generation of the current node to the multi-node server, x is the current power consumption of the current node, a is the first hazard coefficient, b is the second hazard coefficient, and c is the third hazard coefficient.
[0106] Specifically, C is the harm degree of the power generation of the current node to the multi-node server, which is a dimensionless harm degree value, x is the actual power consumption data of the current node collected in real time through the power consumption monitoring module, a is a first harm coefficient, b is a second harm coefficient, and c is a third harm coefficient, which are all fixed parameters calibrated in advance and need to be determined through experimental calibration according to the hardware architecture, power supply redundancy capability, load tolerance limit and other characteristics of the multi-node server. For example, for a high-power consumption node, a is usually set to a positive value (such as 0.002) to amplify the harm weight in the high-power consumption interval, b is a linear adjustment coefficient (such as 0.1) to adapt the harm progression relationship in the medium-power consumption interval, and c is a basic offset coefficient (such as 0.5) to correct the initial harm baseline in the low-power consumption interval.
[0107] In the calculation, the power-on controller first obtains the real-time power consumption x of the current node through the power consumption monitoring module, then calls the harm coefficients a, b and c, and completes the calculation of the value C through the built-in operation unit. After the calculation is completed, the power-on controller compares the value C with the preset harm threshold. The design of the formula can better fit the characteristics of the rapid increase of harm after the power consumption exceeds the standard in the actual scenario, and improve the accuracy and reliability of harm judgment.
[0108] The embodiment adopts a quadratic function formula C=ax 2 +bx+c to calculate the harm degree value, accurately fits the correlation characteristics of power consumption and system harm through a nonlinear model, and can highlight the nonlinear law of rapid increase of harm in the high-power consumption interval compared with the linear calculation method, so that the power-on controller can identify and block the potential threat to the power supply system of the multi-node server and other nodes in time based on the calculation result, and improve the accuracy, adaptability and reliability of power-on safety protection.
[0109] As a specific embodiment of the present application, the working process of the multi-node server power-on system is as shown in Figure 3 , which can include the following steps:
[0110] S101, the system receives a power-on signal, and each node power-on controller starts initialization.
[0111] S102, the current node collects its first power consumption data through the power consumption monitoring module, and obtains the second power consumption data of the adjacent node through the communication interface.
[0112] S103, the power-on controller analyzes all node power consumption data and calculates the power-on condition power consumption of the current node.
[0113] S104, it is judged whether the power-on condition power consumption is ≤ the maximum instantaneous power of the server. If yes, step S105 is executed, and if no, step S103 is executed.
[0114] S105, generate the current node power-on sequence in combination with the first and second power consumption data and the power-on progress information.
[0115] S106, the power-on controller controls the voltage controller to power on the load according to the power-on sequence, and controls the heat dissipation module to start.
[0116] S107, the timing logic indication module feeds back the power-on process through the light group lighting combination, and the power-on controller dynamically adjusts the light group state in combination with the voltage controller feedback information.
[0117] S108, during the power-on process, the power consumption monitoring module collects the current power of the load in real time, and the temperature sensor collects the temperature of the load.
[0118] S109, if the current power of the load exceeds the preset threshold, the temperature exceeds the preset temperature threshold, or the hazard degree value exceeds the preset hazard threshold, the power-on controller controls the voltage controller to stop power-on and triggers the alarm module.
[0119] S110, after the power-on is completed, the power-on result fed back by the load is received, and the subsequent power-on sequence is optimized.
[0120] Therefore, the power supply system of the high-power server proposed in this embodiment has the following technical advantages:
[0121] (1) The dynamic power-on control logic based on node power consumption perception is introduced, which can effectively avoid the problems of startup failure and power module burning caused by power overlimit during power-on of the server.
[0122] (2) Each node has local decision-making ability and does not depend on centralized scheduling, which improves the robustness and fault tolerance of the system operation.
[0123] (3) The timing logic indication module provides clear visual state feedback for real-time display of node grouping state, power supply timing progress and fault prompt.
[0124] (4) The distributed optimization algorithm is fused, the node real-time power consumption and system state are combined for decision-making, the intelligent state control of the voltage controller and the power supply timing judgment are supported, and on-demand and sequential power supply is realized.
[0125] Therefore, the multi-node server power-on system proposed in this embodiment can make collaborative decisions based on real-time power consumption data of each node and real-time power consumption data of adjacent nodes through the power-on controller of each node, dynamically generate a power-on sequence that adapts to the global power consumption state, avoid power superposition impact caused by simultaneous startup of multiple node loads, realize orderly power-on of the load, improve energy utilization efficiency and operation stability, and solve the problem of high power impact during centralized power-on of the load in related technologies.
[0126] The embodiment of the application also provides a control method of a multi-node server power-on system, comprising: in response to a power-on signal, acquiring first power consumption data of a current node by a power consumption monitoring module and acquiring second power consumption data of adjacent nodes by a communication interface; determining whether the current node meets a preset power-on condition according to the first power consumption data and the second power consumption data; in response to the current node meeting the preset power-on condition, generating a power-on sequence of the current node according to the first power consumption data and the second power consumption data; and controlling a voltage controller according to the power-on sequence to power on a load.
[0127] Specifically, when the system receives a power-on signal, the power-on controllers of the nodes start working synchronously, in response to the power-on signal, the first power consumption data of the current node is collected by the power consumption monitoring module, and meanwhile, the power-on controller sends a power consumption query instruction to all adjacent nodes through the communication interface, and the adjacent nodes feed back their current power consumption state as the second power consumption data to the current node after receiving the instruction.
[0128] After the power-on controller acquires the first power consumption data and the second power consumption data of all adjacent nodes, it enters the decision stage, and determines whether the current node meets the preset power-on condition according to the first power consumption data and the second power consumption data. The preset power-on condition can be set based on the overall power supply capacity of the system, the power consumption threshold distribution between nodes, and the principle of avoiding power superposition impact.
[0129] In response to the current node meeting the preset power-on condition, the current node starts power-on, and according to the load characteristics of the first power consumption data and the power consumption redundancy of adjacent nodes reflected by the second power consumption data, the power-on sequence of the current node is dynamically generated. The power-on controller, the voltage controller, the power consumption monitoring module and the timing logic indication module and other devices inside the node need to be powered on in sequence according to the power-on sequence, rather than simultaneously, thereby reducing the power impact.
[0130] According to the power-on sequence, the voltage controller can control the amplitude and rising rate of the output voltage through pulse width modulation and other methods after receiving the instruction of the power-on sequence, to provide smooth power-on power supply for the load, and at the same time, the power consumption monitoring module can continuously collect the actual power consumption data of the load after power-on, and immediately feed back to the power-on controller if an abnormality is found, so that the controller can pause the power-on process or adjust the voltage output to ensure the safety of operation.
[0131] Suppose that a server contains n nodes inside, when the power-on controller of one of the nodes determines that the preset power-on condition is met, the current node starts to be powered on in sequence according to the power-on sequence, then, the adjacent nodes connected to this node calculate whether the total power consumption after joining meets the preset power-on condition according to the power consumption change data transmitted, if it meets, the adjacent nodes start to be powered on, and the power-on control is sequentially promoted to the next adjacent node connected to the adjacent nodes.
[0132] Each node dynamically determines whether the system has sufficient power redundancy based on power acquisition information from neighboring nodes and its own power requirements before performing further power-on operations. This chain-like collaborative control mechanism not only effectively avoids the problem of instantaneous power overload, but also enhances the stability and intelligence of the system.
[0133] The control method for the multi-node server power-on system in this embodiment, through a distributed global collaborative decision-making and dynamic power-on control mechanism, ensures that each node no longer blindly performs power-on operations, but flexibly adjusts its power-on strategy according to its own load characteristics and global power consumption status. This achieves orderly startup of all loads and avoids power surges caused by concentrated power-on, effectively protecting key components such as server power modules and power supply buses. At the same time, it reduces unnecessary peak power reserves and improves energy utilization efficiency and overall system stability.
[0134] Embodiments of this application also provide a multi-node server, such as... Figure 4 As shown, the multi-node server 1000 includes a multi-node server power-on system 100.
[0135] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software 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 specific applications, but such implementations should not be considered beyond the scope of this application.
[0136] The above provides a detailed description of a multi-node server power-on system, control method, and multi-node server provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.
Claims
1. A multi-node server power-on system, characterized in that, The power-on system includes multiple nodes, each node comprising a power-on controller, a communication interface, a voltage controller, a power consumption monitoring module, and a timing logic indication module. The power-on controller is connected to the communication interface, the voltage controller, the power consumption monitoring module, and the timing logic indication module. The power consumption monitoring module is connected to a load, and the voltage controller is connected to the load. Each node is connected to adjacent nodes via the communication interface to form a connected network. The power-on controller is used for: In response to the power-on signal, the first power consumption data of the current node is obtained through the power consumption monitoring module, and the second power consumption data of the neighboring node is obtained through the communication interface; Based on the first power consumption data and the second power consumption data, determine whether the current node meets the preset power-on conditions; In response to the current node meeting the preset power-on conditions, a power-on sequence for the current node is generated based on the first power consumption data and the second power consumption data. The power-on sequence is used to characterize the process in which each power supply branch is gradually powered on in a preset order, voltage amplitude, and time interval during the power-on startup process of the current node. The voltage controller is controlled according to the power-on sequence to power on the load.
2. The multi-node server power-on system according to claim 1, characterized in that, The communication interfaces are connected via a communication bus, and the communication interface includes multiple communication buses. The current node establishes connections with multiple different neighboring nodes through the communication interface.
3. The multi-node server power-on system according to claim 1, characterized in that, The multiple nodes form an undirected interconnected network through the communication interface.
4. The multi-node server power-on system according to claim 1, characterized in that, The power-on controller is also used for: In response to the current node not meeting the preset power-on conditions, the second power consumption data of the adjacent node is continuously acquired through the communication interface; Based on the first power consumption data and the continuously acquired second power consumption data, it is further determined whether the current node meets the preset power-on conditions until the current node meets the preset power-on conditions.
5. The multi-node server power-on system according to claim 4, characterized in that, The second power consumption data includes the power consumption data of all relevant nodes among the plurality of nodes, excluding the current node.
6. The multi-node server power-on system according to claim 5, characterized in that, The power-on controller is specifically used for: The first power consumption data and the second power consumption data are parsed respectively to obtain the current power consumption of each of the plurality of nodes; Calculate the power-on condition power consumption of the current node based on the current power consumption of each node; If the power consumption under the power-on condition is less than or equal to the maximum instantaneous power allowed by the multi-node server, then the current node is determined to meet the preset power-on condition. If the power consumption under the power-on condition is greater than the maximum instantaneous power, then it is determined that the current node does not meet the preset power-on condition.
7. The multi-node server power-on system according to claim 1, characterized in that, The timing logic indicator module includes multiple light groups, which are used to provide feedback on the power-on process of the current node through the lighting combination of the multiple light groups; wherein... The power-on controller is also used to control the plurality of lamp groups according to the power-on sequence.
8. The multi-node server power-on system according to claim 7, characterized in that, The power-on controller is also used for: The system receives feedback information from the voltage controller and controls the plurality of lamp groups according to the feedback information and the power-on sequence.
9. The multi-node server power-on system according to claim 1, characterized in that, The power-on controller is connected to the load, and the power-on controller is further configured to: Receive the power-on result corresponding to the power-on sequence from the load feedback, and optimize the power-on sequence based on the power-on result.
10. The multi-node server power-on system according to claim 1, characterized in that, The second power consumption data includes address information, power consumption information, power-on progress information, and next operation instruction information.
11. The multi-node server power-on system according to claim 10, characterized in that, The power-on controller is also used for: The power-on sequence is optimized based on the power-on progress information and the next operation instruction information.
12. The multi-node server power-on system according to claim 1, characterized in that, Also includes: A temperature sensor is connected to the power-on controller and is disposed on the load to acquire the temperature of the load and send the temperature to the power-on controller.
13. The multi-node server power-on system according to claim 12, characterized in that, The power-on controller is also used for: In response to the temperature exceeding a preset temperature threshold, the voltage controller controls the current node to stop powering on and generates corresponding alarm information.
14. The multi-node server power-on system according to claim 13, characterized in that, Also includes: An alarm module is connected to the power-on controller and is used to receive the alarm information fed back by the power-on controller and provide the alarm information to the target user.
15. The multi-node server power-on system according to claim 1, characterized in that, The power-on controller is also used for: During the process of controlling the voltage controller according to the power-on sequence to power on the load, the current power of the load is obtained in real time through the power consumption monitoring module; In response to the current power exceeding a preset power threshold, the voltage controller controls the current node to stop powering on.
16. The multi-node server power-on system according to claim 1, characterized in that, It also includes a heat dissipation module, which is connected to the power-on controller, wherein the power-on controller is further configured to control the heat dissipation module according to the power-on sequence.
17. The multi-node server power-on system according to claim 1, characterized in that, The power-on controller is also used for: The power consumption monitoring module obtains the real-time power consumption data of the current node and calculates the degree of harm of the power generation of the current node to the multi-node server based on the real-time power consumption data. If the hazard level value is greater than the preset hazard threshold, the voltage controller will control the current node to stop powering on.
18. The multi-node server power-on system according to claim 17, characterized in that, The degree of harm that the current node's power generation poses to the multi-node server is calculated using the following formula: C=ax 2 +bx+c; Where C represents the degree of harm that the power generation of the current node poses to the multi-node server, x represents the current power consumption of the current node, a represents the first harm coefficient, b represents the second harm coefficient, and c represents the third harm coefficient.
19. A control method for a multi-node server power-on system according to any one of claims 1-18, characterized in that, include: In response to the power-on signal, the first power consumption data of the current node is obtained through the power consumption monitoring module, and the second power consumption data of the neighboring node is obtained through the communication interface; Based on the first power consumption data and the second power consumption data, determine whether the current node meets the preset power-on conditions; In response to the current node meeting the preset power-on conditions, a power-on sequence for the current node is generated based on the first power consumption data and the second power consumption data; The voltage controller is controlled according to the power-on sequence to power on the load.
20. A multi-node server, characterized in that, Including the multi-node server power-on system as described in any one of claims 1-18.
Citation Information
Patent Citations
Step-by-step power-on method, device and equipment of server system and storage medium
CN116643640A
DPU card power-on and power-off method, system, device and equipment and storage medium
CN117896191A