Master-slave determination-based multi-power module management method, device, equipment and medium

Abstract

The application provides a multi-power module management method and device based on master-slave determination, equipment and medium. The method is applied to a first power module, the first power module and at least one second power module are arranged in the same cabinet and connected through a communication bus; the method comprises the following steps: in response to a trigger signal, acquiring node broadcast information of each power module in the cabinet through the communication bus; wherein the node broadcast information represents the node attribute and running ability of the power module; determining the arbitration score of each power module according to each node broadcast information; wherein the arbitration score represents the adaptation degree of the power module to undertake the management function; determining the role attribute of the first power module according to each arbitration score; wherein the role attribute is a master node or a slave node, and the power module corresponding to the master node is used for power supply management of the power module corresponding to each slave node. The method can improve the adaptability of the master-slave determination based on the multi-power module and the reliability of the power supply management.

Description

Technical Field

[0001] This application relates to the field of power supply management technology, and in particular to a method, apparatus, equipment and medium for managing multiple power supply modules based on master-slave determination. Background Technology

[0002] In scenarios without a centralized management controller, such as in a data center rack, multiple power modules (e.g., power shelf) are deployed. These power modules typically operate independently to provide power to the equipment within the rack.

[0003] In existing solutions, a specific power module is manually designated as the master node, enabling the master node to manage power distribution and other power supply functions among the power modules within the rack. However, manual configuration relies on manually designating master nodes one by one, which is inefficient and unreliable. Furthermore, manual configuration can only designate a single, fixed master node. If the successfully designated master node fails, the autonomous switching of master node roles between power modules cannot be achieved, thus compromising the continuous execution of the rack's power management functions and resulting in poor reliability of the rack power supply system.

[0004] Therefore, there is an urgent need for a solution that can achieve adaptive master-slave determination in multiple power supply modules to replace manual pre-configuration and improve the reliability of the rack power supply system. Summary of the Invention

[0005] The multi-power module management method, apparatus, device, and medium based on master-slave determination provided in this application can improve the adaptability and reliability of power supply management based on master-slave determination of multi-power modules.

[0006] In a first aspect, embodiments of this application provide a multi-power module management method based on master-slave determination, applied to a first power module, wherein the first power module and at least one second power module are disposed in the same rack and connected via a communication bus, the method comprising:

[0007] In response to a trigger signal, the node broadcast information of each power module in the cabinet is obtained through the communication bus; the trigger signal is used to trigger the power module to perform the master-slave determination process, and the node broadcast information represents the node attributes and operating capabilities of the power module.

[0008] Based on the broadcast information from each node, the arbitration score of each power module is determined; the arbitration score represents the degree to which the power module is suited to assume management functions.

[0009] Based on the arbitration scores, the role attribute of the first power module is determined; the role attribute is either a master node or a slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node.

[0010] If the role attribute of the first power module is the master node and the preset conflict arbitration rule is satisfied, then power supply management is performed on each of the second power modules; wherein, the preset conflict rule represents the rule for determining a unique master node when there are multiple power modules declaring their role attribute as master nodes.

[0011] Optionally, as described above, the node broadcast information of each power module in the rack is obtained via the communication bus, including:

[0012] The system broadcasts node broadcast information from the first power module via the communication bus and receives node broadcast information from each of the second power modules.

[0013] Receive node broadcast information from each power module in the cabinet.

[0014] Optionally, as described above, the arbitration score for each power module is determined based on the broadcast information from each node, including:

[0015] For each power module, a set of indicators is extracted based on the node broadcast information; the set of indicators includes one or more of the following: slot priority, health status, power supply capacity, communication quality, firmware level, and fault status.

[0016] Determine the weight coefficient corresponding to each indicator in the indicator set, and determine the arbitration score of the power module based on each indicator in the indicator set and its corresponding weight coefficient.

[0017] Arbitration scores were obtained for each power module.

[0018] Optionally, as described above, the role attributes of the first power module are determined based on each arbitration score, including:

[0019] If the arbitration score of the first power module is greater than the arbitration scores of each of the second power modules, then the role attribute of the first power module is determined to be the main node.

[0020] If the arbitration score of the first power module is less than the arbitration score of any second power module, then the role attribute of the first power module is determined to be a slave node.

[0021] Optionally, as described above, if the first power module's role attribute is a master node and it meets the preset conflict arbitration rules, then power supply management is performed on each of the second power modules, including:

[0022] If the role attribute of the first power module is the master node, it sends identity declaration information to each of the second power modules; wherein, the identity declaration information is used to indicate that the role attribute of the power module sending the identity declaration information is the master node;

[0023] If an identity declaration message is received from the second power module, the adjustment parameters corresponding to each power module are determined based on the node broadcast information of each power module; wherein, the adjustment parameters represent one or more of the slot identifier, node identifier, and communication timestamp.

[0024] Based on the indicator set of the first power module and the corresponding adjustment parameters, the latest arbitration score of the first power module is determined, and based on the indicator set of the second power module and the corresponding adjustment parameters, the latest arbitration score of the second power module is determined.

[0025] If the latest arbitration score of the first power module is greater than the latest arbitration score of the second power module, then the preset conflict arbitration rules are met, the latest identity declaration information is sent to each second power module, and power supply management is performed on each second power module.

[0026] Optionally, as described above, the role attribute of the first power module is the master node; the method further includes:

[0027] Based on a preset transmission period, heartbeat information is sent to each second power module; wherein, the heartbeat information is used by the second power module to output a trigger signal when it determines that it has not received heartbeat information within the preset detection period.

[0028] Optionally, as described above, the role attribute of the first power module is the master node; the method further includes:

[0029] The system sends status acquisition commands to each second power module via the communication bus and receives status data from each second power module; the status data represents the real-time operating status of the power module.

[0030] The status data of each power module in the cabinet is processed into structured data.

[0031] Based on a pre-defined digital model of the power supply domain, a power supply strategy is determined according to structured data; the power supply strategy is used to allocate power and control redundancy for each power module.

[0032] Secondly, embodiments of this application provide a multi-power module management device based on master-slave determination, applied to a first power module, wherein the first power module and at least one second power module are disposed in the same rack and connected via a communication bus; the device includes:

[0033] The information acquisition unit is used to acquire the node broadcast information of each power module in the cabinet through the communication bus in response to the trigger signal; wherein, the trigger signal is used to trigger the power module to perform the master-slave determination process, and the node broadcast information represents the node attributes and operating capabilities of the power module.

[0034] The scoring determination unit is used to determine the arbitration score of each power module based on the broadcast information of each node; the arbitration score represents the degree of suitability of the power module to assume management functions.

[0035] The master-slave determination unit is used to determine the role attribute of the first power module based on each arbitration score; wherein, the role attribute is master node or slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node.

[0036] The conflict resolution unit is used to manage the power supply of each second power module if the role attribute of the first power module is a master node and the preset conflict arbitration rule is met; wherein, the preset conflict rule represents the rule for determining a unique master node when there are multiple power modules declaring their role attributes as master nodes.

[0037] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0038] The memory stores instructions that the computer executes;

[0039] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0040] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0041] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0042] The multi-power module management method, apparatus, device, and medium based on master-slave determination provided in this application embodiment are applied to a first power module. The first power module and at least one second power module are arranged in the same rack and connected through a communication bus. The first power module obtains the node broadcast information of each power module in the rack through the communication bus in response to a trigger signal. Further, based on the node broadcast information, the arbitration score of each power module is determined. Further, based on the arbitration score, the role attribute of the first power module is determined. The trigger signal is used to trigger the power module to perform the master-slave determination process. The node broadcast information represents the node attribute and operating capability of the power module. The arbitration score represents the adaptability of the power module to undertake management functions. The role attribute is a master node or a slave node. The power module corresponding to the master node is used to perform power supply management for the power modules corresponding to each slave node. This application's multi-power module management method based on master-slave determination automatically initiates the master-slave determination process upon response to a trigger signal, replacing the existing method of manually pre-configuring master nodes. This avoids the tedious operation of manually assigning master nodes one by one, improving the efficiency of master-slave determination. By determining the arbitration score based on node broadcast information, the master-slave determination can comprehensively consider the node attributes and operational capabilities of each power module, achieving adaptive election of master nodes and improving the rationality of master node determination. By determining role attributes based on the arbitration score, each power module can autonomously determine its master-slave role according to its own compatibility with other nodes, realizing the ability to autonomously switch master node roles among power modules and ensuring the continuous execution of rack power supply management functions. The method of this application improves the adaptability of master-slave determination based on multi-power modules and the reliability of power supply management. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0044] Figure 1 A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 1 ;

[0045] Figure 2 A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 2 ;

[0046] Figure 3 A flowchart illustrating the power supply management method provided in this application;

[0047] Figure 4 This is a schematic diagram of the system topology provided in this application;

[0048] Figure 5A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 3 ;

[0049] Figure 6 A schematic diagram of the structure of the multi-power module management device based on master-slave determination provided in this application;

[0050] Figure 7 A schematic diagram of the structure of the electronic device provided in this application.

[0051] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0052] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0053] It should be noted that the multi-power module management method, apparatus, equipment and medium based on master-slave determination in this application can be used in the field of power supply management technology, or in any field other than the field of power supply management technology. The application field of the power supply management technology in this application is not limited.

[0054] In artificial intelligence training or inference operations, rack load can fluctuate significantly depending on the operating status of devices such as graphics processing units (GPUs) and central processing units (CPUs). Therefore, power modules (such as power shelves) not only need to provide basic power supply, but also need to have certain collaborative management capabilities to maintain the stability and continuity of rack-level power supply.

[0055] In existing solutions, multi-power modules are typically managed by operating independently, manually pre-setting fixed master-slave relationships, or simply collecting basic status data via a bus. The basic idea is usually that a pre-defined master node assumes management responsibilities, while other power modules cooperate according to predetermined relationships, or each module outputs power independently based on its local status.

[0056] While the above methods can meet basic power supply requirements, they lack a unified status awareness and role determination mechanism for all power modules within the rack. Master nodes often rely on static configurations, making it difficult to adjust based on changes in node attributes and operational capabilities. When the health status, power supply capacity, or communication status of a power module changes, the existing methods typically cannot reflect this in the master-slave relationship in a timely manner, leading to a mismatch between management roles and actual operating conditions.

[0057] Especially when the master node is abnormal, communication fluctuates, or the cabinet load changes rapidly, problems such as the master node being unable to effectively assume management responsibilities, slave nodes lacking clear collaboration objects, and power supply management continuity may occur. This can lead to uneven power distribution, delayed fault takeover, and increased operation and maintenance complexity, making it difficult to meet the requirements of high power density cabinets for adaptive master-slave determination and power supply management.

[0058] In light of this, improving the adaptability and management reliability of the master-slave determination process within a multi-power module architecture in the same rack has become a pressing technical problem. Specifically, a solution is urgently needed that enables adaptive master-slave determination within multiple power modules to replace manual pre-configuration and improve the reliability of the rack power supply system.

[0059] This application's multi-power module management method based on master-slave determination automatically initiates the master-slave determination process upon response to a trigger signal, replacing the existing method of manually pre-configuring master nodes. This avoids the tedious operation of manually assigning master nodes one by one, improving the efficiency of master-slave determination. By determining the arbitration score based on node broadcast information, the master-slave determination can comprehensively consider the node attributes and operational capabilities of each power module, achieving adaptive election of master nodes and improving the rationality of master node determination. By determining role attributes based on the arbitration score, each power module can autonomously determine its master-slave role according to its own compatibility with other nodes, realizing the ability to autonomously switch master node roles among power modules and ensuring the continuous execution of rack power supply management functions. The method of this application improves the adaptability of master-slave determination based on multi-power modules and the reliability of power supply management.

[0060] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0061] Figure 1 A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 1 The method is applied to a first power module, which is housed in the same rack as at least one second power module and connected via a communication bus. The method includes:

[0062] S101. In response to the trigger signal, the node broadcast information of each power module in the cabinet is obtained through the communication bus; wherein, the trigger signal is used to trigger the power module to perform the master-slave determination process, and the node broadcast information represents the node attributes and operating capabilities of the power module.

[0063] S102. Determine the arbitration score of each power module based on the broadcast information of each node; wherein, the arbitration score represents the degree of suitability of the power module to assume management functions;

[0064] S103. Determine the role attribute of the first power module based on each arbitration score; wherein, the role attribute is master node or slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node.

[0065] S104. If the role attribute of the first power module is a master node and the preset conflict arbitration rule is satisfied, then power supply management is performed on each second power module; wherein, the preset conflict rule represents the rule for determining a unique master node when multiple power modules declare their role attributes as master nodes.

[0066] The first power module can refer to any power module (e.g., a Power Shelf) within the rack that requires master-slave role determination. In a master-slave determination process, the first power module is the active party executing the method described in this application. It participates in arbitration scoring calculations by obtaining node broadcast information from all power modules within the rack, including itself, and ultimately determines its own role attributes. The first power module and the second power module can have identical hardware structures, both possessing independent power supply capabilities, status acquisition capabilities, communication capabilities, and arbitration calculation capabilities.

[0067] The second power supply can refer to any power module other than the first power module within the rack. The second power module is located in the same rack as the first power module and interconnected via the same communication bus. During the master-slave determination process, the second power module also broadcasts its own node broadcast information and participates in the arbitration scoring calculation. There can be one or more second power modules, depending on the rack's power configuration requirements. For example, an Artificial Intelligence (AI) rack can be configured with two, four, or more power modules, each deployed in a different slot within the rack.

[0068] A server rack can refer to the physical framework structure used to house IT equipment such as servers, storage devices, network equipment, and power modules. In this application, a server rack specifically refers to a power supply unit that deploys multiple power modules, such as an AI rack, server rack, or network equipment rack in a data center. The rack contains multiple slots for installing various power modules, and the slots are connected for power and communication via backplanes or cables. The rack may also include a cooling system (e.g., fans), environmental sensors (e.g., temperature sensors), and interfaces for connecting to external management systems.

[0069] A communication bus refers to the physical transmission medium and communication protocol used to enable data communication between power modules within a cabinet. The communication bus can be a Power Management Bus (PMBus), a bidirectional two-wire synchronous serial (Inter-Integrated Circuit, I2C) bus, or a Controller Area Network-BUS (CAN-BUS) bus. The communication bus is configured to support broadcast or multicast modes, allowing messages sent by any power module to be received by other power modules within the same cabinet.

[0070] In step S101, the trigger signal can refer to any signal used to initiate the master-slave determination process. For example, the trigger signal may include, but is not limited to, a power-on trigger signal, a timed trigger signal, and a fault detection trigger signal. Specifically, the power-on trigger signal can be a signal generated by the control unit after the power module completes its startup initialization, used to initiate the master-slave determination process after the cabinet is first powered on or the power module restarts; the timed trigger signal can be a signal generated by an internal timer of the power module according to a preset period, used to periodically trigger the master-slave determination process to dynamically evaluate the status of each power module; the fault detection trigger signal can be a signal generated when the slave node detects an anomaly in the master node, used to initiate a new master-slave determination process to achieve master node switching when the master node fails.

[0071] Node broadcast information refers to the set of information representing the node attributes and operational capabilities of each power module, sent by each power module to other power modules in the rack via the communication bus. For example, node broadcast information includes, but is not limited to, node identifier (which can be represented as Node ID), slot identifier (which can be represented as Slot ID), power capacity (which can be represented as PowerCapability), health status (which can be represented as Health Status), and firmware version (which can be represented as FirmwareVersion). Specifically, the node identifier is a globally unique identifier for the power module within the rack, used to distinguish different power modules; the slot identifier is the physical slot number of the power module in the rack, used to identify the physical installation location of the power module; the power capacity is the actual output power currently available to the power module, used to reflect the power module's ability to handle loads; the health status is the overall operational health status of the power module, used to assess whether the power module currently meets the conditions for normal operation; and the firmware version is the version number of the firmware program currently running on the power module, used to identify the software functional level of the power module.

[0072] In practice, after responding to the trigger signal, the first module can perform the following operations:

[0073] First, the first power module broadcasts its own node broadcast information via the communication bus, enabling other power modules in the rack to obtain the node attributes and operational capabilities of the first power module. Then, the first power module listens for and receives node broadcast information from other power modules (i.e., the second power module) in the rack via the communication bus. Finally, it obtains the node broadcast information of all power modules in the rack.

[0074] It should be understood that the specific type and source of the trigger signal can be configured according to the actual operational needs of the cabinet. For example, when the cabinet is first powered on, the power-on trigger signal is the primary one; during stable operation, a timed trigger signal can be configured to achieve periodic master-slave relationship evaluation; in scenarios with high reliability requirements, a fault detection trigger signal can be enabled to achieve rapid failover. Each trigger signal can be used individually or in combination, such as configuring both a power-on trigger signal and a timed trigger signal simultaneously.

[0075] In step S102, the arbitration score characterizes the suitability of the power module to assume management functions, that is, the suitability of the power module as a master node to perform cabinet power supply management functions.

[0076] In specific implementation, the first power module can extract key fields from the broadcast information of each node obtained in step S101 through parsing and processing, construct a topology information table according to a unified data format, and store it in local storage (such as memory or cache), and then calculate the arbitration score of each power module according to the preset arbitration scoring rules.

[0077] The preset arbitration scoring rules may include, but are not limited to, weighted scoring rules (e.g., normalizing at least one field in the node broadcast information and summing it according to preset weight coefficients to obtain the arbitration score) and priority sorting rules (e.g., determining priority according to the slot identifier in ascending order, with the smaller the slot number, the higher the priority, and directly using the priority sorting result as the arbitration basis). The specific implementation form of the arbitration scoring rules is not restricted here.

[0078] It should be understood that the first power module can independently calculate the arbitration score of each power module, or each power module can independently calculate it based on the same node broadcast information to ensure the consistency of the arbitration results. In one possible implementation, each power module calculates independently based on the same node broadcast information and the same arbitration scoring rules. Since the input data and calculation rules are consistent, the arbitration scores calculated by each power module are the same, thus achieving master-slave consistency without the need for an additional arbitration coordination mechanism.

[0079] In step S103, the role attribute is either a master node or a slave node. The power module corresponding to the master node is used to manage the power supply of the power modules corresponding to each slave node, which may include, but is not limited to, management functions such as power budget allocation, redundancy strategy management, status monitoring, and fault diagnosis. The power modules corresponding to the slave nodes accept the management of the master node, such as executing power scheduling commands issued by the master node and periodically reporting their own status information to the master node.

[0080] In practice, the first power module can compare the arbitration scores of all power modules in the rack and determine the power module with the highest arbitration score as the master node. Specifically, if the first power module has the highest arbitration score, then the first power module is designated as the master node; otherwise, the first power module is designated as the slave node.

[0081] In step S104, the preset conflict rule is used to determine the unique master node from multiple candidate master nodes when multiple power modules simultaneously declare the master node identity during the master-slave determination process.

[0082] In practice, after determining its role as the master node, the first power module sends an identity declaration message via the communication bus. Simultaneously, the first power module monitors the communication bus for identity declaration messages from other power modules. If no other identity declaration messages are received, it is determined that the preset conflict arbitration rules are met, and the first power module is confirmed as the sole master node, performing power supply management for each second power module. If other identity declaration messages are received, it is determined that a master node identity conflict exists, triggering a conflict resolution process.

[0083] It should be understood that the specific implementation method of the preset conflict arbitration rules can be flexibly set according to the rack configuration requirements. For example, a hierarchical comparison method can be adopted (comparing slot identifier, node identifier, and communication timestamp in sequence), or a weighted calculation method can be adopted (incorporating the adjustment parameters as an additional item into the arbitration score calculation). No restrictions are placed on the specific implementation method here.

[0084] This embodiment provides a multi-power module management method based on master-slave determination. By automatically initiating the master-slave determination process in response to a trigger signal, it replaces the existing method of manually pre-configuring master nodes, avoiding the tedious operation of manually assigning master nodes one by one and improving the efficiency of master-slave determination. By determining the arbitration score based on node broadcast information, the master-slave determination can comprehensively consider the node attributes and operational capabilities of each power module, achieving adaptive election of the master node and improving the rationality of master node determination. By determining role attributes based on the arbitration score, each power module can autonomously determine its master-slave role according to its own compatibility with other nodes, realizing the autonomous switching capability of master node roles among power modules and ensuring the continuous execution of the rack power supply management function. The method of this application improves the adaptability of master-slave determination based on multi-power modules and the reliability of power supply management.

[0085] Figure 2 A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 2 , Figure 2 Implementation examples in Figure 1 Based on the embodiments, the multi-power module management method based on master-slave determination is further explained, such as... Figure 2 As shown, the method includes:

[0086] S201. In response to the trigger signal, broadcast the node broadcast information of the first power module through the communication bus, and receive the node broadcast information broadcast by each second power module; obtain the node broadcast information of each power module in the cabinet.

[0087] Specifically, after detecting a trigger signal, the control unit of the first power module (e.g., a microcontroller unit (MCU)) encapsulates the node broadcast information into a data frame of a preset format and sends it via the communication bus using the broadcast address. Simultaneously, the first power module activates a listening mode to receive node broadcast information from other power modules (i.e., the second power module) within a preset time window. After receiving the information, the first power module parses and stores the broadcast information from each node in its local memory (e.g., RAM or cache) for subsequent arbitration scoring.

[0088] It is understandable that by automatically broadcasting node information through the communication bus and receiving broadcast node information from other nodes, mutual discovery and status awareness of each power module in the cabinet can be achieved without manually configuring node information or specifying master-slave relationships, thus providing a data foundation for adaptive master-slave determination.

[0089] S202. For each power module, extract a set of indicators based on the node broadcast information; wherein, the set of indicators includes one or more of the following: slot priority, health status, power supply capacity, communication quality, firmware level, and fault status.

[0090] The indicator set can refer to a set of parameters of multiple dimensions used to evaluate the adaptability of the power module to assume management functions. Each indicator is extracted from the corresponding field in the node broadcast information or calculated. The values ​​of each indicator in the indicator set can be mapped to a unified numerical range (e.g., 0 to 1) through normalization processing to achieve the uniformity of the dimensions and comparability of each indicator.

[0091] For the slot priority metric in the metric set, slot priority refers to the positional priority of the power module in the physical slot of the rack, as represented by the slot identifier field in the node broadcast information. Its value can be determined as follows: the smaller the slot ID, the higher the slot priority. For example, slot 0 has a higher slot priority than slot 1. For instance, the slot priority metric for slot 0 can be represented as 100%, and the slot priority metric for slot 1 can be represented as 75%.

[0092] For the health status indicator in the indicator set, health status can refer to the overall operational health status of the power module as represented by the health status field in the node broadcast information. It reflects whether the power module currently has the conditions for normal operation. Its value can be calculated and normalized by weighting the input voltage, output voltage, output current, temperature, and power supply unit (PSU) status collected by the power module in the node broadcast information. For example, the value is 100% when the input voltage is stable, the output voltage is normal, the output current is not overloaded, the temperature is below the threshold, and all PSUs are online; the value is 70% when any PSU is offline; the value is 50% when the temperature exceeds the preset alarm threshold; and the value is 30% when the input voltage is abnormal.

[0093] For the power supply capacity indicator in the indicator set, power supply capacity can refer to the actual output power currently available to the power module as represented by the power supply capacity field in the node broadcast information. It is used to reflect the power module's ability to bear the load. Its value can be the ratio of the currently available power supply capacity to the rated power and normalized. For example, when the rated power is 3000 watts and the currently available power is 2250 watts, the value is 75%.

[0094] For the communication quality indicator in the indicator set, the communication quality can be calculated from the underlying communication statistics of the communication bus (such as packet loss rate, communication delay, and number of retransmissions). It is used to reflect the reliability of communication between the power module and other nodes. Its value can be calculated and normalized based on the communication delay and packet loss rate. For example, when the delay is less than 2 milliseconds and the packet loss rate is 0, the value is 100%. For every 1 millisecond increase in delay or every 1% increase in packet loss rate, the value of communication quality decreases according to a preset gradient.

[0095] For the firmware level indicator in the indicator set, the firmware level can refer to the functional level corresponding to the firmware version number currently running on the power module, as represented by the firmware version field in the node broadcast information. It is used to identify the software functional level of the power module. Its value can be obtained by comparing the firmware version number with the highest version number in the rack and normalizing it. For example, the firmware version number can be converted into a value through a preset mapping rule and compared with the value corresponding to the highest version number. The difference ratio between the two can be calculated. The closer the firmware version number is to the latest version number, the higher the firmware level indicator value. When running the latest version, the value is 100%.

[0096] For the fault status indicator in the indicator set, the fault status can refer to the historical and current fault status of the power module represented by the fault record field in the node broadcast information, reflecting the reliability record of the power module. Its value can be calculated and normalized based on the number of historical faults and the current fault flag. For example, the value is 0 when there are no historical faults and no current faults, 25% when there is one historical fault and no current faults, and 100% when there is a current fault.

[0097] It should be understood that the specific definitions and formats of each field in the node broadcast information can be flexibly set according to the communication protocol and rack configuration, and no restrictions are placed on the specific data structure of the fields here; the indicator set can include some or all of the above six indicators, and the specific combination can be flexibly adjusted according to rack configuration requirements and management precision requirements. The value selection method and normalization mapping rule of each indicator can be preset or dynamically adjusted according to the actual application scenario, and no restrictions are placed on the specific form of the value selection method and normalization mapping rule here.

[0098] S203. Determine the weight coefficient corresponding to each indicator in the indicator set, and determine the arbitration score of the power module based on each indicator in the indicator set and its corresponding weight coefficient.

[0099] The weighting coefficients represent the importance of each indicator in the arbitration score, and the sum of the weighting coefficients is 1.

[0100] In one possible implementation, the weight coefficients can be determined by reading the default weight coefficients from a preset configuration file; dynamically adjusting them according to the rack operation scenario (e.g., increasing the weight coefficients corresponding to the health status in a high-reliability scenario); or receiving weight coefficient configurations remotely issued by the host computer management system.

[0101] In one possible implementation, the arbitration score of the power module can be obtained by multiplying the values ​​of each indicator in the indicator set (specifically, the normalized values) by their corresponding weight coefficients and then summing the results.

[0102] It should be understood that the method of determining the arbitration score is not limited to weighted summation. Other mathematical models can also be used, such as machine learning-based scoring models, as long as they can comprehensively reflect the degree of suitability of the power module to assume management functions.

[0103] S204. Obtain the arbitration score for each power module.

[0104] Specifically, the first power module determines the arbitration score for each power module (including the first power module itself and each second power module) in the manner described above, thereby obtaining the arbitration score of all power modules in the cabinet, which serves as the basis for determining the role attributes in subsequent steps.

[0105] It is understandable that by extracting multi-dimensional indicators and configuring corresponding weight coefficients for weighted calculation, the arbitration score of each power module can be objectively quantified. This can comprehensively evaluate the suitability of each power module to undertake management functions, thereby providing data basis for the scientific and fair election of master nodes and realizing the rationality and reliability of adaptive election of master nodes.

[0106] S205. If the arbitration score of the first power module is greater than the arbitration scores of each of the second power modules, then the role attribute of the first power module is determined to be the master node.

[0107] Specifically, the first power module compares its own arbitration score with the arbitration scores of each of the second power modules one by one. If the arbitration score of the first power module is greater than the arbitration scores of all the second power modules, the first power module determines its role as a master node and prepares to execute the master node-related operations.

[0108] It should be understood that the arbitration score of the first power module is greater than the arbitration scores of each of the second power modules, that is, the arbitration score of the first power module is greater than the arbitration scores of all the second power modules. If there are cases where the arbitration scores are the same and all are the highest scores, an output trigger signal (used to trigger the re-execution of the master-slave determination process) or a conflict elimination arbitration process (which can determine a unique master node after further comparing the priority of slot identifier, node identifier or communication timestamp) can be triggered to ensure the uniqueness of the master node election.

[0109] S206. If the arbitration score of the first power module is less than the arbitration score of any second power module, then the role attribute of the first power module is determined to be a slave node.

[0110] Specifically, if the arbitration score of at least one second power module is greater than that of the first power module, the first power module determines its role as a slave node and prepares to receive management instructions from the master node.

[0111] It should be understood that if the arbitration score of the first power module is less than the arbitration score of any second power module, that is, if the arbitration score of the first power module is lower than the arbitration score of at least one second power module, then the first power module will give up the master node competition, enter the slave node state, wait for and receive the identity declaration information sent by the power module with the highest arbitration score, so as to establish a master-slave management relationship.

[0112] It is understandable that determining role attributes through quantitative comparison of arbitration scores ensures that the master node is the power module with the highest degree of compatibility in undertaking management functions within the rack, and that the number of master nodes is unique. This can avoid power management chaos caused by mismatched master node capabilities or multiple management, thereby ensuring the efficiency and stability of rack power management.

[0113] In one optional implementation, if the first power module's role is that of a master node and satisfies a preset conflict arbitration rule, then power supply management is performed on each of the second power modules, including:

[0114] S207. If the role attribute of the first power module is the master node, it sends identity declaration information to each of the second power modules; wherein, the identity declaration information is used to indicate that the role attribute of the power module sending the identity declaration information is the master node.

[0115] Among them, the identity declaration information can refer to the broadcast message (which can be represented as a Master Claim message) used to declare the identity of the power module master node. For example, it can include information such as the declared node identifier, arbitration score value, and timestamp.

[0116] Specifically, the first power module sends identity declaration information to each of the second power modules via the communication bus in a broadcast or unicast manner, declaring its own master node identity.

[0117] By sending identity declaration information, the identity of the master node can be synchronized to all power modules in the rack, enabling each slave node to clearly identify the management entity, thereby establishing a stable master-slave management mode.

[0118] It should be understood that the power module corresponding to the master node is used to manage the power supply of the power modules corresponding to each slave node, including but not limited to management functions such as power budget allocation, redundancy strategy management, status monitoring, fault diagnosis, and power limiting command issuance.

[0119] S208. If an identity declaration information is received from the second power module, the adjustment parameters corresponding to each power module are determined according to the node broadcast information of each power module; wherein, the adjustment parameters represent one or more of the slot identifier, node identifier and communication timestamp.

[0120] In this context, while the first power module is designated as the master node, it still receives identity declaration information from the second power module. This indicates that multiple power modules simultaneously identify themselves as master nodes (i.e., the arbitration scores of the first and second power modules are equal), leading to a conflict in master node identities. Therefore, it is necessary to introduce adjustment parameters to further differentiate priorities in order to eliminate the conflict.

[0121] Adjustment parameters can refer to additional distinguishing parameters used to eliminate arbitration conflicts. Specifically, adjustment parameters represent one or more of the following: slot identifier, node identifier, and communication timestamp, all of which can be determined by extracting the corresponding fields from the node broadcast information.

[0122] Among them, the communication timestamp can refer to the time mark when the identity declaration information is sent, which is used to record the order in which each power module declares the identity of the master node.

[0123] S209. Based on the indicator set of the first power module and the corresponding adjustment parameters, determine the latest arbitration score of the first power module, and based on the indicator set of the second power module and the corresponding adjustment parameters, determine the latest arbitration score of the second power module.

[0124] Specifically, the adjustment parameters can be incorporated into the indicator set as additional indicators to form a new indicator set, or used as additional weights to adjust the weights of each indicator in the original indicator set, thereby recalculating the arbitration score to obtain the latest arbitration score. Specifically, the adjustment parameters represent one or more of the following: slot identifier, node identifier, and communication timestamp.

[0125] In one possible implementation, the order of use of the parameters in the adjustment parameters can be slot identifier, node identifier, and communication timestamp, respectively. For example, firstly, the slot identifier is used as the adjustment parameter to determine the latest arbitration score of the first power module based on its indicator set and the corresponding slot identifier, and the latest arbitration score of the second power module is determined based on its indicator set and the corresponding slot identifier. If the latest arbitration score of the first power module is still equal to the latest arbitration score of the second power module, then the node identifier is used as the adjustment parameter to determine the latest arbitration score of the first power module based on its indicator set and the corresponding node identifier, and the latest arbitration score of the second power module is determined based on its indicator set and the corresponding node identifier. If the latest arbitration score of the first power module is still equal to the latest arbitration score of the second power module, then the communication timestamp is used as the adjustment parameter to determine the latest arbitration score of the first power module based on its indicator set and the corresponding communication timestamp, and the latest arbitration score of the second power module is determined based on its indicator set and the corresponding communication timestamp.

[0126] In one possible implementation, in the aforementioned steps, the arbitration score determined solely based on the power module's set of indicators can satisfy scoring formula one:

[0127] ;

[0128] in, Indicates the arbitration score. Indicates slot priority. This represents the weight coefficient corresponding to the slot priority. Indicates health status. The weighting coefficients represent the corresponding health status. Indicates power supply capacity. This represents the weighting coefficient corresponding to the power supply capacity. Indicates communication quality. This represents the weighting coefficient corresponding to communication quality. Indicates firmware level. This represents the weighting coefficient corresponding to the firmware level. Indicates the fault status. These represent the weighting coefficients corresponding to the fault states. All weighting coefficients are positive numbers.

[0129] When there is a conflict in the identity of the master node, taking the slot identifier as an example, the method for determining the latest arbitration score of the first power module and the latest arbitration score of the second power module can be as follows: Based on the above scoring formula one, the slot identifier is used as an additional weight. The slot identifier as an additional weight coefficient is multiplied by the weight coefficient of the slot priority to obtain scoring formula two:

[0130] ;

[0131] in, This indicates that the slot identifier has been incorporated as the latest arbitration score after parameter adjustments. This indicates the additional weight coefficient corresponding to the slot identifier.

[0132] In one possible implementation, a mapping table is pre-set between slot identifiers and additional weight coefficients, with a one-to-one correspondence between the slot identifiers and the additional weight coefficients. The additional weight coefficient corresponding to the slot identifier is retrieved from the mapping table. This mapping table can be pre-configured in local memory for each power module at the factory or during deployment to support fast lookup and real-time arbitration calculation.

[0133] Furthermore, if the latest arbitration score of the first power module and the latest arbitration score of the second power module are still equal when the adjustment parameter is the slot identifier, the node identifier can be used as the adjustment parameter, and the steps for determining the latest arbitration scores of the first and second power modules can be repeated: Based on the above scoring formula one, the node identifier is used as an additional indicator, for example, it can satisfy scoring formula three:

[0134] ;

[0135] in, This indicates that the node identifier has been incorporated as the latest arbitration score after parameter adjustments. This represents the normalized value of the node identifier. This represents the weight coefficient corresponding to the node identifier.

[0136] In one possible implementation, a mapping table between node identifiers and normalized values ​​is pre-defined, with a one-to-one correspondence between node identifiers and normalized values. The normalized value corresponding to a node identifier is retrieved from the mapping table. This mapping table can be pre-configured in the local memory of each power module at the factory or during deployment, or it can be uniformly configured by a host computer management system. Correspondingly, the weight coefficient corresponding to the node identifier can be a preset fixed value.

[0137] Furthermore, if the latest arbitration score of the first power module and the latest arbitration score of the second power module are still equal when the adjustment parameter is the node identifier, the communication timestamp can be used as the adjustment parameter to re-determine the latest arbitration scores of the first and second power modules: Based on the above scoring formula one, the node identifier is used as an additional indicator, for example, it can satisfy scoring formula three:

[0138] ;

[0139] in, This indicates the latest arbitration score after incorporating the communication timestamp as an adjusted parameter. This represents the normalized value of the communication timestamp. This represents the weighting coefficient corresponding to the communication timestamp.

[0140] In one possible implementation, a pre-defined mapping rule between communication timestamps and normalized values ​​is established. Communication timestamps are mapped to decreasing normalized values ​​according to the order in which identity declaration information is sent. For example, the power module that sends identity declaration information first corresponds to a normalized value of 1, and subsequent power modules are mapped with decreasing values ​​based on time intervals; the earlier the communication timestamp, the larger the normalized value. Correspondingly, the weighting coefficient corresponding to the communication timestamp can be a pre-defined fixed value.

[0141] It should be understood that the calculation rules for the latest arbitration score of the first power module and the latest arbitration score of the second power module meet the same calculation standard, that is, the first power module and the second power module adopt the same scoring formula two (or scoring formula three, scoring formula four) and the same additional weight coefficient mapping rules to ensure the fairness and consistency of the conflict elimination arbitration.

[0142] Specifically, both the first power module and the second power module independently calculate the latest arbitration score based on their respective sets of indicators and the same adjustment parameter processing rules. Since the calculation rules are consistent, the calculation results are comparable, thus enabling the objective determination of a unique master node.

[0143] S210. If the latest arbitration score of the first power module is greater than the latest arbitration score of the second power module, then the preset conflict arbitration rules are met, the latest identity declaration information is sent to each second power module, and power supply management is performed on each second power module.

[0144] Specifically, if the latest arbitration score of the first power module is greater than that of the second power module, meaning the first power module wins the conflict resolution arbitration and has a higher master node priority, then it broadcasts its latest identity declaration information to each of the second power modules through the communication bus to confirm the legitimacy of its master node identity.

[0145] If the latest arbitration score of the first power module is lower than that of the second power module, the first power module will voluntarily relinquish its master node status, switch to slave node status, and receive the latest identity declaration information sent by the second power module.

[0146] In one possible implementation, the latest identity statement information may include the updated arbitration score (i.e., the latest arbitration score) and adjustment parameters, which are used to prove the uniqueness and legitimacy of the master node's identity to all power modules in the rack.

[0147] It is understandable that by introducing adjustment parameters to handle arbitration conflicts, and comparing them again in the case of the same arbitration score to distinguish the priority of each power module as the master node, it can be ensured that there is still a clear basis for adjudication when the arbitration scores of the first power module and the second power module are the same, thereby ensuring the uniqueness and certainty of the master node election and realizing the robustness and reliability of the master-slave determination process.

[0148] In one optional implementation, the first power module is designated as the master node. Considering that a master node failure needs to be promptly detected by the slave nodes to trigger a new master-slave determination process and ensure the continuity of power supply management, the above steps may further include:

[0149] S211. Based on a preset transmission period, send heartbeat information to each second power module; wherein, the heartbeat information is used by the second power module to output a trigger signal when it determines that it has not received heartbeat information within the preset detection period.

[0150] The heartbeat information refers to the periodic broadcast messages indicating the master node's operational status. Specifically, the heartbeat information may include the master node identifier and a heartbeat timestamp. For example, the sending period can be configured to 1 second, 2 seconds, or 5 seconds, and the detection period can be three times the sending period (e.g., a sending period of 2 seconds and a detection period of 6 seconds). If the second power module (in this embodiment, the power module with the role of a slave node) does not receive heartbeat information within the detection period, it determines that the master node is abnormal, generates a trigger signal, and initiates a new master-slave determination process.

[0151] The fault trigger signal can be the trigger signal from step S201 above, used to trigger the re-execution of the master-slave determination process to achieve automatic switching of the master node. Specifically, after detecting a heartbeat timeout, the second power module outputs a trigger signal to its own control unit. In response to the trigger signal, the control unit re-executes the master-slave determination process to elect a new master node.

[0152] It should be understood that the heartbeat detection mechanism of each power module is configured in the same way, so the conditions and standards for outputting the trigger signal are the same. This ensures that each slave node makes the same judgment on the fault of the master node, and avoids conflicts caused by multiple nodes triggering the master-slave determination process at the same time due to differences in judgment.

[0153] It is understandable that by periodically sending heartbeat information and configuring the corresponding detection mechanism, the slave node can be triggered to automatically start a new master-slave determination process when the master node fails. This allows for rapid switching of the master node without manual intervention, ensuring the continuous execution of the cabinet power supply management function.

[0154] In this embodiment, based on the above embodiments, on the one hand, by extracting multi-dimensional indicators and configuring weight coefficients for arbitration scoring, the problem that existing manual configuration of master nodes cannot comprehensively consider the actual capabilities of each node is solved; on the other hand, through identity declaration, conflict elimination and heartbeat mechanisms, the uniqueness of master node election and the stability of master-slave relationship are further guaranteed; thereby achieving the effect of adaptive master node election, while improving the automation level and reliability of rack power supply management.

[0155] Figure 3 This is a flowchart illustrating the power supply management method provided in this application. Figure 3 This embodiment, based on the aforementioned embodiments, specifically the multi-power module management method based on master-slave determination, explains how to manage the power supply of a power module whose role attribute is the master node (optionally, the first power module's role attribute is the master node). Figure 3 As shown, the method includes:

[0156] S301. Send status acquisition commands to each second power module via the communication bus and receive status data from each second power module; wherein, the status data represents the real-time operating status of the power module.

[0157] Among them, the status acquisition command can refer to the control message used to request each power module to report its real-time operating status.

[0158] When the role attribute of the first power module is determined to be the master node, the role attribute of each second power module is the slave node. The first power module periodically sends status acquisition commands to each second power module through the communication bus in order to receive status data from each second power module.

[0159] Status data can refer to the set of real-time operating status information reported by each power module after responding to the status acquisition command. Specifically, status data can be acquired in real time by the power module through its local status acquisition module, and then transmitted back to the first power module in response to the status acquisition command of the first power module.

[0160] In one possible implementation, the status acquisition command can be sent in a polling manner (i.e., the master node queries each slave node in turn) or a broadcast manner (i.e., the master node broadcasts the status acquisition command uniformly, and each slave node actively reports). The specific method can be flexibly selected according to the communication bus bandwidth and real-time requirements.

[0161] S302. Perform structured processing on the status data of each power module in the cabinet to obtain structured data.

[0162] In one possible implementation, the structured processing can be as follows: parsing the original format of each state data, mapping and converting fields according to a unified data model (such as a preset relevant digital model); cleaning abnormal data (such as removing values ​​that are obviously outside the reasonable range); interpolating or marking missing data; organizing according to dimensions such as node identifier, timestamp, and state type, and finally generating a structured power supply domain state table.

[0163] For example, structured data may include, but is not limited to, PSU status, power budget, redundancy status, server power requirements, temperature summary status, leakage summary status, and fan summary status. Specifically, PSU status may refer to the status of each PSU in the power module (e.g., online, offline, or faulty) and its input voltage, output voltage, input current, and output current values; power budget may refer to the current available power and allocated power of the power module; redundancy status may refer to the current redundancy mode (e.g., N+1 mode or N+N mode) and the online status of redundant power modules; server power requirements may refer to the current power consumption and peak power requirements of each server; temperature summary status may refer to the temperature summary information of key temperature measurement points within the rack; leakage summary status may refer to the leakage detection status of the liquid cooling system within the rack; and fan summary status may refer to the operating status and speed information of the cooling fans within the rack.

[0164] It is understandable that structured data facilitates unified analysis and decision-making by the master node, and supports rapid power supply strategy generation and fault location.

[0165] S303. Based on the preset power supply domain digital model, determine the power supply strategy according to structured data; the power supply strategy is used to perform power allocation and redundancy control for each power module.

[0166] The preset power supply domain digital model can refer to a pre-constructed mathematical or logical model used to describe the operating status and constraints of the cabinet power supply system. For example, by inputting structured data into the preset power supply domain digital model, an evaluation result reflecting the current cabinet power supply status and the corresponding power supply strategy can be obtained.

[0167] A power supply strategy refers to a control scheme generated by the master node based on the power domain digital model and real-time structured data, used to guide the operation of each power module and manage the power supply of the rack. The power supply strategy can be expressed as a power allocation table, a redundancy control instruction set, a fault isolation instruction, a power limit request, etc., and is sent to each slave node for execution via the communication bus in the form of data frames or instruction sequences. For example, the power allocation table includes the output power limit of each power module, the power budget allocation ratio of each server, and dynamic adjustment rules, correspondingly implementing power allocation for each power module. The redundancy control instruction set includes redundancy mode configuration (e.g., N+1 mode or N+N mode), redundancy module enable / disable instructions, and fault switching logic, correspondingly implementing redundancy control for each power module.

[0168] It should be understood that the specific construction method of the power supply domain digital model and the algorithm for determining the power supply strategy can be flexibly selected according to the complexity of the rack, the management accuracy requirements, and the availability of computing resources. No restrictions are placed on the specific mathematical form of the model or the specific implementation method of the algorithm. Furthermore, the power supply strategy can be updated periodically or triggered by events to adapt to dynamic changes in rack load.

[0169] In this embodiment, based on the aforementioned embodiments, by performing structured processing on real-time status data and determining power supply strategies based on a preset power supply domain digital model, the power allocation and redundancy control of each power module are guided, thereby achieving unified perception and intelligent decision-making of rack-level power supply status. This can optimize power allocation efficiency, improve the flexibility of redundancy management, and thus enhance the reliability of power supply management.

[0170] To more intuitively demonstrate the rack system architecture and component connection relationships of this application, Figure 4 The system topology diagram provided in this application is as follows: Figure 4 As shown, in the system architecture of the cabinet (which can be an AI RACK), there are multiple Power Shelves, namely Power Shelf (1), Power Shelf (2), Power Shelf (3), and Power Shelf (4).

[0171] In response to the trigger signal, each Power Shelf initiates the master-slave determination process, and finally determines Power Shelf (1) as the master node (representing the rack-level management node), while Power Shelf (2), Power Shelf (3), and Power Shelf (4) are all slave nodes.

[0172] Each PowerShelf has a consistent built-in structure, including an Ethernet module, a node information module, a status acquisition module, an arbitration module, a heartbeat module, a master claim module, and a power control module. The Ethernet module communicates with the host computer management system; the node information module stores node broadcast information, such as node identifier, slot identifier, firmware version, and power level; the status acquisition module collects real-time operating status data of the power modules, such as input voltage, output voltage, output current, temperature, and PSU status; the arbitration module determines arbitration scores and role attributes; the heartbeat module periodically sends or receives heartbeat information to detect master node failures; the master claim module sends or receives identity claim information to declare or confirm the master node's identity; and the power control module performs power management functions such as power scheduling, redundancy management, and power limiting.

[0173] Each Power Shelf can connect to a switch and interact with the host computer management system via Ethernet. The switch connects to the host computer to enable remote monitoring and unified management of the rack-level power supply status.

[0174] Simultaneously, each Power Shelf can receive broadcast information or management commands from the Rack Shared Management Bus via the communication bus or external signal. The Rack Shared Management Bus is connected to the server management interface module, fan control module, leak detection module, and status indicator module via the communication bus or external signal. Specifically, the server management interface module is used to enable the Power Shelf to obtain server operating status and send control commands to the server; the fan control module is used to detect fan status and send fan status information to the Power Shelf; the leak detection module is used to detect leaks and send leak status information to the Power Shelf; and the status indicator module is used to display the operating status of the AI ​​Rack based on the control commands from the Power Shelf.

[0175] It should be understood that the Rack Shared Management Bus can use various communication protocols such as I2C, PMBus, or Ethernet. The specific protocol selection can be determined based on the communication requirements and compatibility requirements of each module within the rack. Furthermore, the specific number and configuration of each functional module can be flexibly adjusted according to the rack size and business needs; therefore, no restrictions are placed on the number of modules or their configuration methods.

[0176] Figure 5A flowchart illustrating the multi-power module management method based on master-slave determination provided in this application. Figure 3 , Figure 5 Based on the foregoing embodiments, this example provides another illustration of the multi-power module management method based on master-slave determination, applied to a specific power shelf, such as... Figure 5 As shown, the method includes:

[0177] S501, System Power On.

[0178] S502, Initialization.

[0179] This initialization indicates Power Shelf initialization.

[0180] S503: Broadcast node information of its own node through the communication bus.

[0181] S504: Receive broadcast node information from other nodes via the communication bus.

[0182] S505. Based on the broadcast information of each node, construct the topology information table.

[0183] S506. Determine the arbitration score for all nodes within the cabinet.

[0184] S507. Determine whether the arbitration score of its own node is the highest arbitration score.

[0185] S508. If the arbitration score of its own node is not the highest arbitration score, it waits to receive identity declaration information.

[0186] This identity claim information can be represented as a Master Claim message.

[0187] S509. If the arbitration score of its own node is the highest arbitration score, then send identity declaration information.

[0188] S510. Determine whether identity statement information has been received from other nodes.

[0189] S511. If no identity declaration information is received from other nodes, then become the master node.

[0190] S512. If identity statement information is received from other nodes, conflict arbitration is performed and step S509 is re-executed.

[0191] The logic of conflict arbitration is consistent with that of steps S208 to S210 above, and will not be repeated here.

[0192] Based on the above master-slave-based multi-power module management method, the implementation logic for a specific scenario is given below:

[0193] The four Power Shelves are interconnected via I2C as follows: Slot 0: Power Shelf A, Slot 1: Power Shelf B, Slot 2: Power Shelf C, and Slot 3: Power Shelf D.

[0194] After system startup, each Power Shelf powers on, initializes, and broadcasts node information. Power Shelf A, identified as Slot 0, is in good health, has sufficient power supply, excellent communication quality, the highest firmware level, and no fault records. It receives the highest arbitration score after arbitration scoring and becomes the master node. Power Shelf B, Power Shelf C, and Power Shelf D, as slave nodes, receive the identity declaration information sent by Power Shelf A to confirm the master-slave relationship.

[0195] Power Shelf A establishes a power domain status table to uniformly manage the power budget of the entire cabinet.

[0196] During operation, if Power Shelf A fails, other nodes (Power Shelf B, Power Shelf C, and Power Shelf D) detect Heartbeat timeouts (e.g., failing to receive heartbeat information from Power Shelf A three consecutive times), triggering a re-execution of the arbitration process. Each node rebroadcasts its node broadcast information and recalculates its arbitration score. Power Shelf B, with the highest arbitration score, becomes the new master node.

[0197] The system determines the fault type and recoverability of the power Shelf A's status table. If the fault in Power Shelf A is temporary (e.g., communication interruption, software anomaly) and the power domain status table data can be recovered from backup storage or cache, the power domain status table established by Power Shelf A is synchronized. If the fault in Power Shelf A is permanent (e.g., hardware damage, power loss) and the power domain status table data cannot be recovered, the power domain status table is re-established. Power Shelf B then continues to perform power domain management. The entire process requires no manual intervention.

[0198] Figure 6 A schematic diagram of the multi-power module management device based on master-slave determination provided in this application is shown below. Figure 6 As shown, the multi-power module management device 60 based on master-slave determination includes: an information acquisition unit 601, a scoring determination unit 602, a master-slave determination unit 603, and a conflict elimination unit 604.

[0199] The information acquisition unit 601 is used to acquire the node broadcast information of each power module in the cabinet through the communication bus in response to the trigger signal; wherein, the trigger signal is used to trigger the power module to perform the master-slave determination process, and the node broadcast information represents the node attributes and operating capabilities of the power module.

[0200] The scoring determination unit 602 is used to determine the arbitration score of each power module based on the broadcast information of each node; wherein, the arbitration score represents the degree of adaptability of the power module to assume management functions;

[0201] The master-slave determination unit 603 is used to determine the role attribute of the first power module based on each arbitration score; wherein, the role attribute is a master node or a slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node.

[0202] The conflict resolution unit 604 is used to perform power supply management on each of the second power modules if the role attribute of the first power module is a master node and the preset conflict arbitration rule is satisfied; wherein, the preset conflict rule represents the rule for determining a unique master node when there are multiple power modules declaring their role attribute as master nodes.

[0203] In an optional example, the information acquisition unit 601 is also used for:

[0204] The system broadcasts node broadcast information from the first power module via the communication bus and receives node broadcast information from each of the second power modules.

[0205] Receive node broadcast information from each power module in the cabinet.

[0206] In an optional example, the scoring determination unit 602 is also used for:

[0207] For each power module, a set of indicators is extracted based on the node broadcast information; the set of indicators includes one or more of the following: slot priority, health status, power supply capacity, communication quality, firmware level, and fault status.

[0208] Determine the weight coefficient corresponding to each indicator in the indicator set, and determine the arbitration score of the power module based on each indicator in the indicator set and its corresponding weight coefficient.

[0209] Arbitration scores were obtained for each power module.

[0210] In an optional example, the master-slave determination unit 603 is also used for:

[0211] If the arbitration score of the first power module is greater than the arbitration scores of each of the second power modules, then the role attribute of the first power module is determined to be the main node.

[0212] If the arbitration score of the first power module is less than the arbitration score of any second power module, then the role attribute of the first power module is determined to be a slave node.

[0213] In an optional example, the role attribute of the first power module is master node; the multi-power module management device 60 based on master-slave determination also includes an identity declaration unit, which is used for:

[0214] If the role attribute of the first power module is the master node, it sends identity declaration information to each of the second power modules; wherein, the identity declaration information is used to indicate that the role attribute of the power module sending the identity declaration information is the master node;

[0215] If an identity declaration message is received from the second power module, the adjustment parameters corresponding to each power module are determined based on the node broadcast information of each power module; wherein, the adjustment parameters represent one or more of the slot identifier, node identifier, and communication timestamp.

[0216] Based on the indicator set of the first power module and the corresponding adjustment parameters, the latest arbitration score of the first power module is determined, and based on the indicator set of the second power module and the corresponding adjustment parameters, the latest arbitration score of the second power module is determined.

[0217] If the latest arbitration score of the first power module is greater than the latest arbitration score of the second power module, then the preset conflict arbitration rules are met, the latest identity declaration information is sent to each second power module, and power supply management is performed on each second power module.

[0218] In an optional example, the role attribute of the first power module is master node; the multi-power module management device 60 based on master-slave determination also includes a heartbeat sending unit, which is used for:

[0219] Based on a preset transmission period, heartbeat information is sent to each second power module; wherein, the heartbeat information is used by the second power module to output a trigger signal when it determines that it has not received heartbeat information within the preset detection period.

[0220] In an optional example, the role attribute of the first power module is master node; the multi-power module management device 60 based on master-slave determination also includes a power supply management unit, which is used for:

[0221] The system sends status acquisition commands to each second power module via the communication bus and receives status data from each second power module; the status data represents the real-time operating status of the power module.

[0222] The status data of each power module in the cabinet is processed into structured data.

[0223] Based on a pre-defined digital model of the power supply domain, a power supply strategy is determined according to structured data; the power supply strategy is used to allocate power and control redundancy for each power module.

[0224] The multi-power module management device based on master-slave determination in this embodiment can execute the technical solutions in the above method. Its specific implementation process and technical principles are the same, and will not be repeated here.

[0225] Figure 7 A schematic diagram of the structure of the electronic device provided in this application, such as... Figure 7 As shown, the electronic device 70 provided in this embodiment includes at least one processor 701 and a memory 702. Optionally, the electronic device 70 further includes a communication component 703. The processor 701, memory 702, and communication component 703 are connected via a bus 704.

[0226] In a specific implementation, at least one processor 701 executes computer execution instructions stored in memory 702, causing at least one processor 701 to perform the above-described method.

[0227] The specific implementation process of processor 701 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0228] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0229] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0230] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0231] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0232] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0233] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as magnetic disks or optical disks.

[0234] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A multi-power module management method based on master-slave determination, characterized in that, The method applies to each power module in a server rack, the server rack including multiple power modules connected via a communication bus; the method includes: In response to a trigger signal, the node broadcast information of each power module in the cabinet is obtained through the communication bus; wherein, the trigger signal is used to trigger the power module to perform a master-slave determination process, and the node broadcast information is a set of information used to characterize the node attributes and operating capabilities of the power module; Based on the broadcast information of each node and the preset arbitration scoring rules, the arbitration score of each power module is determined; wherein, the arbitration score represents the degree of suitability of the power module to assume management functions; wherein, each power module independently executes the steps of determining the arbitration score of each power module based on the same node broadcast information and the same arbitration scoring rules. Based on the arbitration scores, the role attribute of the power module executing the method is determined; wherein, the role attribute is a master node or a slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node; If the role attribute of the power module executing the method is the master node and the preset conflict arbitration rule is satisfied, then power supply management is performed on the remaining power modules; wherein, the preset conflict arbitration rule represents the rule for determining a unique master node when there are multiple power modules declaring their role attribute as master nodes.

2. The method according to claim 1, characterized in that, The step of obtaining node broadcast information of each power module in the cabinet through the communication bus includes: The power module executing the method broadcasts node broadcast information via the communication bus, and receives node broadcast information broadcast by other power modules. The node broadcast information of each power module in the cabinet is obtained.

3. The method according to claim 1, characterized in that, The step of determining the arbitration score of each power module based on the broadcast information of each node and the preset arbitration scoring rules includes: For each power module, an indicator set is extracted based on the node broadcast information; wherein, the indicator set includes one or more of the following: slot priority, health status, power supply capacity, communication quality, firmware level, and fault status. Determine the weight coefficient corresponding to each indicator in the indicator set, and determine the arbitration score of the power module based on each indicator in the indicator set and its corresponding weight coefficient. Arbitration scores are obtained for each of the power modules.

4. The method according to claim 1, characterized in that, The determination of the role attribute of the power module performing the method based on each of the arbitration scores includes: If the arbitration score of the power module executing the method is greater than the arbitration score of the other second power modules, then the role attribute of the power module executing the method is determined to be the master node; If the arbitration score of the power module executing the method is less than the arbitration score of any other power module, then the role attribute of the power module executing the method is determined to be the slave node.

5. The method according to claim 3, characterized in that, If the power module executing the method has the role attribute of the master node and meets the preset conflict arbitration rules, then power supply management is performed on the remaining power modules, including: If the role attribute of the power module executing the method is the master node, it sends identity declaration information to the other power modules; wherein, the identity declaration information is used to indicate that the role attribute of the power module sending the identity declaration information is the master node; If an identity declaration message is received from the second power module, the adjustment parameters corresponding to each power module are determined based on the node broadcast information of each power module; wherein, the adjustment parameters represent one or more of slot identifier, node identifier, and communication timestamp; the second power module is at least one of the remaining power modules; Based on the indicator set of the power module executing the method and the corresponding adjustment parameters, the latest arbitration score of the power module executing the method is determined, and based on the indicator set of the second power module and the corresponding adjustment parameters, the latest arbitration score of the second power module is determined. If the latest arbitration score of the power module executing the method is greater than the latest arbitration score of the second power module, then it is determined that the preset conflict arbitration rule is met, the latest identity declaration information is sent to the remaining power modules, and power supply management is performed on the remaining power modules.

6. The method according to claim 1, characterized in that, The role attribute of the power module executing the method is the master node; the method further includes: Based on a preset transmission period, heartbeat information is sent to the other power modules; wherein, the heartbeat information is used by the other power modules to output the trigger signal when they determine that they have not received heartbeat information within the preset detection period.

7. The method according to any one of claims 1-6, characterized in that, The role attribute of the power module executing the method is the master node; the method further includes: The system sends status acquisition commands to the other power modules via the communication bus and receives status data from the other power modules; wherein the status data represents the real-time operating status of the power modules. The status data of each power module in the cabinet is processed into structured data. Based on a preset power supply domain digital model, a power supply strategy is determined according to the structured data; the power supply strategy is used to perform power allocation and redundancy control on each of the power modules.

8. A multi-power module management device based on master-slave determination, characterized in that, A power supply module applied to each power supply module in a server rack, the server rack including multiple power supply modules connected via a communication bus; the device includes: An information acquisition unit is used to acquire node broadcast information of each power module in the cabinet through the communication bus in response to a trigger signal; wherein, the trigger signal is used to trigger the power module to perform a master-slave determination process, and the node broadcast information is a set of information used to characterize the node attributes and operating capabilities of the power module. The scoring determination unit is used to determine the arbitration score of each power module based on the broadcast information of each node and the preset arbitration scoring rules; wherein the arbitration score represents the degree of adaptability of the power module to assume management functions; wherein each power module independently performs the step of determining the arbitration score of each power module based on the same node broadcast information and the same arbitration scoring rules. The master-slave determination unit is used to determine the role attribute of the power module of the device based on the arbitration scores; wherein the role attribute is a master node or a slave node, and the power module corresponding to the master node is used to manage the power supply to the power modules corresponding to each slave node. The conflict resolution unit is used to manage the power supply of other power modules if the role attribute of the power module of the device is the master node and a preset conflict arbitration rule is met; wherein, the preset conflict arbitration rule is a rule for determining a unique master node when multiple power modules declare their role attribute as master nodes.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images