A system and intelligent supercomputing cluster for cluster firmware upgrades

Abstract

The application provides a cluster firmware upgrading system and an intelligent supercomputing cluster, and the system comprises: an intelligent PDU having a relay switch and a plurality of power supply sockets; an ARM MCU chip is arranged in the intelligent PDU, and a GPIO pin is connected to the relay switch; an RJ45 network interface is connected to the ARM MCU chip, configured to be connected to a switch and to each device node in the cluster through the switch; a storage card is connected to the ARM MCU chip, and a Master Linux system is pre-installed in the storage card; after the intelligent PDU is powered on, the Master Linux system is started and runs on the ARM MCU chip; a USB interface is connected to the ARM MCU chip, and the USB interface is configured to transmit control software, a refreshing program, a firmware image file and a refreshing tool required for firmware refreshing into the Master Linux system running in the ARM MCU chip. By using the scheme of the application, the problem of low efficiency caused by manual intervention in the firmware upgrading and maintenance process can be solved, and the automation and batch processing of the entire firmware refreshing process can be realized.

Description

Technical Field

[0001] This invention relates to the field of computers, and more specifically to a system for cluster firmware upgrades and an intelligent supercomputing cluster. Background Technology

[0002] With the development of technologies such as artificial intelligence and big data, intelligent computing centers have become a crucial infrastructure in today's society. Essentially, an intelligent computing center is a computing cluster composed of tens of thousands of supercomputing devices. Against the backdrop of Moore's Law gradually losing its effectiveness, humanity's demand for computing power continues to grow rapidly. To address this issue, parallel computing and heterogeneous computing have developed rapidly. Taking the field of artificial intelligence as an example, GPUs provide the most versatile and widely used computing resources for the entire industry, and GPU systems are a typical example of heterogeneous computing systems.

[0003] A significant characteristic of AI model training is the need for ultra-dense computing power and high-throughput inter-device communication. To meet this requirement, heterogeneous computing systems have evolved into increasingly complex systems. Taking GPUs as an example, the most advanced system currently is the NVIDIA HGX system, which integrates eight GPU chips within a single compute node, while relying on multiple NVSWITCH chips and PCIe SWITCH chips to provide high-throughput I / O capabilities. To manage such a complex system, HGX integrates management chips such as FPGA, EROT, and HMC. The presence of these chips presents challenges for firmware upgrades and maintenance, as the device needs to be powered off after each upgrade command to activate the newly flashed firmware image.

[0004] When a target computer needs a version upgrade, the upgrade of each component is completed through the target computer's own operating system and IPMI interface. When some devices require a power outage, the power cord is manually disconnected for the operation. Because some components require an AC power outage to take effect after the upgrade, and there are specific requirements for the upgrade sequence of each component, multiple power outages and restarts are required during the upgrade process. Therefore, the upgrade process requires manual intervention and cannot achieve batch automation of the upgrade process. Summary of the Invention

[0005] In view of this, the purpose of this invention is to propose a system for cluster firmware upgrade and an intelligent supercomputing cluster. By using the technical solution of this invention, the problem of low efficiency due to the need for human intervention in the firmware upgrade and maintenance process can be solved, and the entire firmware refresh process can be automated and batch-processed.

[0006] To achieve the above objectives, according to one aspect of the present invention, a system for cluster firmware upgrade is provided, comprising:

[0007] A smart PDU (Reachctrl Power, also known as a smart PDU, IP power supply, intelligent power, or rack-mounted power distribution unit, which combines power distribution and management functions) has a relay switch and multiple power sockets.

[0008] ARM (microprocessor) MCU (microcontroller unit) chip: The ARM MCU chip is located inside the smart PDU, and the GPIO pins of the ARM MCU chip are connected to the relay switch.

[0009] The RJ45 network interface connects to the ARM MCU chip and is configured to connect to a switch and then to various device nodes in the cluster.

[0010] The memory card is connected to the ARM MCU chip. The Master Linux system is pre-installed in the memory card. The memory card is configured to start the Master Linux system and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on.

[0011] The USB interface connects to the ARM MCU chip and is configured to transfer the control software, flashing program, firmware image file, and flashing tools required for firmware flashing to the Master Linux system running on the ARM MCU chip.

[0012] According to one embodiment of the present invention, the GPIO pin of the ARM MCU chip outputs a high-level signal or a low-level signal to the relay switch of the smart PDU so that the relay switch controls the on / off state of the power jack based on the received level signal.

[0013] According to one embodiment of the present invention, the Master Linux system also runs a RAMOS server, which integrates DHCP service and NFS service. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing service.

[0014] According to one embodiment of the present invention, the control software is configured to determine whether the device node needs to be powered off based on the status information written by each device node. When it is determined that the device node needs to be powered off, the corresponding level signal is sent to the relay switch through the GPIO pin of the ARM MCU chip.

[0015] According to one embodiment of the present invention, the flashing program, firmware image file and flashing tool required for flashing the firmware are all stored in a shared folder provided by the NFS service.

[0016] According to another aspect of the present invention, an intelligent supercomputing cluster is also provided, the intelligent supercomputing cluster including a cluster firmware upgrade system, the cluster firmware upgrade system comprising:

[0017] A smart PDU features a relay switch and multiple power jacks.

[0018] The ARM MCU chip is located inside the smart PDU, and its GPIO pins are connected to a relay switch.

[0019] The RJ45 network interface connects to the ARM MCU chip and is configured to connect to a switch and then to various device nodes in the cluster.

[0020] The memory card is connected to the ARM MCU chip. The Master Linux system is pre-installed in the memory card. The memory card is configured to start the Master Linux system and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on.

[0021] The USB interface connects to the ARM MCU chip and is configured to transfer the control software, flashing program, firmware image file, and flashing tools required for firmware flashing to the Master Linux system running on the ARM MCU chip.

[0022] According to one embodiment of the present invention, the GPIO pin of the ARM MCU chip outputs a high-level signal or a low-level signal to the relay switch of the smart PDU so that the relay switch controls the on / off state of the power jack based on the received level signal.

[0023] According to one embodiment of the present invention, the Master Linux system also runs a RAMOS server, which integrates DHCP service and NFS service. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing service.

[0024] According to one embodiment of the present invention, the control software is configured to determine whether the device node needs to be powered off based on the status information written by each device node. When it is determined that the device node needs to be powered off, the corresponding level signal is sent to the relay switch through the GPIO pin of the ARM MCU chip.

[0025] According to one embodiment of the present invention, the flashing program, firmware image file and flashing tool required for flashing the firmware are all stored in a shared folder provided by the NFS service.

[0026] The present invention has the following beneficial technical effects: The cluster firmware upgrade system provided in this embodiment of the invention, by setting up a smart PDU, which has a relay switch and multiple power sockets; an ARM MCU chip, which is set inside the smart PDU, with its GPIO pins connected to the relay switch; an RJ45 network interface, which is connected to the ARM MCU chip and configured to connect to a switch and then to each device node in the cluster; a memory card, which is connected to the ARM MCU chip and pre-installed with a Master Linux system, configured to start the Master Linux system and run it on the ARM MCU chip after the smart PDU is powered on; and a USB interface, which is connected to the ARM MCU chip and configured to transfer the control software, flashing program, firmware image file, and flashing tools required for firmware flashing to the Master Linux system running on the ARM MCU chip, can solve the problem of low efficiency caused by the need for manual intervention during firmware upgrade and maintenance, and can realize the automation and batch processing of the entire firmware flashing process. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of a cluster firmware upgrade system according to an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of the software framework of a cluster firmware upgrade system according to an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram illustrating the connection between a cluster firmware upgrade system and a device according to an embodiment of the present invention;

[0031] Figure 4 This is a schematic diagram of a firmware upgrade process according to an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to specific examples and the accompanying drawings.

[0033] Based on the above objectives, a first aspect of the embodiments of the present invention provides an embodiment of a system for cluster firmware upgrade. Figure 1 The diagram shown is a schematic of the system.

[0034] like Figure 1 As shown, the system may include:

[0035] A smart PDU features a relay switch and multiple power jacks. It includes multiple standard 16A high-power AC jacks for powering node devices.

[0036] It also includes an ARM MCU chip, which is located within the intelligent PDU. The GPIO pins of the ARM MCU chip are connected to a relay switch. The PDU's built-in ARM MCU chip runs a Master Linux operating system, within which also runs a RAMOS server, as well as the control software and flashing program required for firmware flashing. The RAMOS server is a software service running on the Linux operating system, capable of booting a node and directly loading a complete Linux system into the target node's memory and running it, even without an operating system installed. The RAMOS service also integrates DHCP and NFS services; DHCP provides IP addresses for the network interfaces on the device node, and NFS provides shared file services. The main flashing program, firmware image file, and flashing tools running on the target node are all located in the NFS shared folder. The ARM chip has GPIO pins that can output high and low level signals. The GPIO pins are connected to a relay switch, which controls the on / off state of a 16A high-power AC jack. Within the Linux operating system, the register values ​​of the GPIO pins can be read and written by the program, thereby controlling the pin level signals. The control software is located within the Master Linux system. The main function of the control software is to determine when to power off the relay based on the status information written to each device node in the NFS file system, and then execute the AC power-off operation.

[0037] It also includes an RJ45 network interface, which connects to the ARM MCU chip and is configured to connect to a switch and then to the various device nodes in the cluster.

[0038] The memory card connects to the ARM MCU chip and pre-installs the Master Linux system. The memory card is configured to boot and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on. Alternatively, the memory card can be a microSD card with a pre-installed Linux system, called Master Linux, which automatically boots and runs on the ARM chip when the PDU is powered on.

[0039] The USB interface connects to the ARM MCU chip and is configured to transfer the control software, flashing program, firmware image file, and flashing tools required for firmware flashing to the Master Linux system running on the ARM MCU chip.

[0040] By integrating these functional modules, this invention deploys all the software tools required for firmware upgrades within the PDU, providing the target node with all the resources needed, including power, operating system, flashing tools and images, and process control programs. This eliminates the need for manual intervention in the upgrade process, enabling automated batch operations.

[0041] The technical solution of this invention can solve the problem of low efficiency caused by the need for human intervention in the firmware upgrade and maintenance process, and can realize the automation and batch processing of the entire firmware refresh process.

[0042] like Figure 1 As shown, the PDU incorporates an ARM MCU chip, which serves as the main controller for the entire PDU. External devices connected to this chip include an SD card, an RJ45 interface, a USB interface, and a relay switch. After the high-voltage power supply is connected to the PDU, it first passes through the relay switch and then connects to each socket. The power on / off state of the sockets is controlled by the relay switch, and the switching state of the relay is controlled by signals from the ARM MCU chip.

[0043] like Figure 2As shown, the software system mainly includes the RAMOS server and the control program required for firmware flashing. RAMOS is a technology that allows an operating system to be installed into the target computer's memory and run directly without an operating system installed. It consists of DHCP, PXE, and NFS services. The DHCP service is used to assign IP addresses to each node device, including BMC IPs and operating system IPs. PXE is used to boot the node device and decompress the file system into memory. The NFS service provides shared file storage services for the node devices. The firmware flashing control program mainly controls the process according to the required sequence and flow during the actual flashing process, ensuring that the entire flashing process executes according to the predefined steps. This process also involves capturing the flashing status of all target nodes, judging and handling errors, and controlling the switching of relays.

[0044] In a preferred embodiment of the present invention, the GPIO pin of the ARM MCU chip outputs a high-level signal or a low-level signal to the relay switch of the smart PDU so that the relay switch controls the on / off state of the power jack based on the received level signal.

[0045] In a preferred embodiment of the present invention, the Master Linux system also runs a RAMOS server, which integrates DHCP service and NFS service. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing service.

[0046] In a preferred embodiment of the present invention, the control software is configured to determine whether the device node needs to be powered off based on the status information written by each device node. When it is determined that the device node needs to be powered off, the corresponding level signal is sent to the relay switch through the GPIO pin of the ARM MCU chip.

[0047] In a preferred embodiment of the present invention, the flashing program, firmware image file and flashing tool required for flashing the firmware are all stored in a shared folder provided by the NFS service.

[0048] Example

[0049] As shown in Table 1 below, to upgrade the firmware from version A to version B, each device needs to be upgraded individually: GPU, NVSWITH, PCIE SWITCH, FPGA, and CEC.

[0050] Table 1 Firmware Version Comparison

[0051] GPU 92.00.36.00.04 92.00.81.00.01 NVSWITCH 92.10.18.00.01 92.10.18.00.02 PCIe Switch v1.2f v3.1f FPGA v2.A5 v3.03 CEC v3.9 v4.0

[0052] The manufacturer recommends the following refresh order:

[0053] ① Upgrade CEC firmware;

[0054] ② Power off and restart the node device to activate the new CEC firmware;

[0055] ③ Upgrade the PCIe switch firmware;

[0056] ④ Upgrade GPU firmware;

[0057] ⑤ Upgrade the NVSWITCH firmware;

[0058] ⑥ Reset the node device to activate the newly flashed firmware;

[0059] ⑦ Upgrade FPGA firmware;

[0060] ⑧ Power off and restart the node device using AC power to activate the new FPGA firmware.

[0061] like Figure 3 As shown, the power supply cable of the device node is plugged into the socket of the intelligent PDU, and the PDU provides the power required for the device to operate. The network interface of the device node is connected to the RJ45 interface on the intelligent PDU through a switch, and the DHCP service in the intelligent PDU provides IP addresses to all devices. The workflow of the entire system is as follows: Figure 4 As shown, after the intelligent PDU is powered on, the Master Linux system in the PDU starts, the relay switch closes, the node device is powered on, and the DHCP service assigns IPMIs to the node. The IP control system powers on the node via IPMI commands. PXE boots the node device to run the RAMOS system, which is mounted on the NFS shared file system. The main flashing program retrieves the local firmware upgrade list and determines if the upgrade is complete. If so, the process ends; otherwise, it flashes the firmware device by device in a predefined order. After each device is flashed, a check is performed to determine if a power outage or reboot is required. If not, the program continues to the next device. If a reboot is required, the main flashing program issues a reboot command to the device. After rebooting, the process returns to the PXE booting step to run the RAMOS system and continues flashing other firmware devices according to the same steps. If a reboot is required, the device status is written to the NFS directory using the IP identifier. The Master obtains information from each node and issues a disconnect command to the relay. After a certain delay (ideally 20 seconds), the Master issues a close command to the relay, which closes. The process then returns to the PXE booting step to run the RAMOS system and continues flashing other firmware devices according to the same steps until all firmware updates are complete.

[0062] To enable the Master Linux system to obtain the status of each device node, information synchronization needs to be achieved through NFS shared files. Specifically, the main refresh program on each device node creates a folder in the Nodes directory of the NFS shared file based on its own IP address and writes the current execution progress to the status.json file in real time.

[0063]

[0064]

[0065]

[0066]

[0067] The refresh control program in the Master Linux system obtains the refresh progress of each device by reading the status.json file corresponding to each node. The names of each data segment in status.json correspond to the process that the firmware actually needs to follow for refresh. Here, name refers to the refresh step; timeout refers to the timeout period set for the current step, which is used for fault tolerance. When the refresh process encounters an exception and cannot be executed normally, the Master control program can stop the operation according to the timeout period; status indicates the execution status of the current step, done means that it has been completed, wait means that the step has not yet been reached, and start means that the step needs to be executed.

[0068] The control program in the Master Linux system continuously polls the status.json of each node in the Nodes folder. When the status of a step with the name "ac" on a certain node changes to "start", it continues to poll and waits for the status of the same step on other devices to change to "start". Then, the control program changes the corresponding status to "done" and controls the relay to disconnect via a command.

[0069] This invention primarily addresses the inefficiency caused by the need for manual intervention during firmware upgrades and maintenance of some high-end intelligent computers. By integrating the power supply module and the software module required for firmware flashing, the software module can determine whether power is needed based on the status information of the node device and control the power supply to the node device via relays, thus automating and batch processing the entire process.

[0070] Based on the above objectives, a second aspect of the embodiments of the present invention proposes an intelligent supercomputing cluster, the intelligent supercomputing cluster including a cluster firmware upgrade system, the cluster firmware upgrade system comprising:

[0071] A smart PDU features a relay switch and multiple power jacks.

[0072] The ARM MCU chip is located inside the smart PDU, and its GPIO pins are connected to a relay switch.

[0073] The RJ45 network interface connects to the ARM MCU chip and is configured to connect to a switch and then to various device nodes in the cluster.

[0074] The memory card is connected to the ARM MCU chip. The Master Linux system is pre-installed in the memory card. The memory card is configured to start the Master Linux system and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on.

[0075] The USB interface connects to the ARM MCU chip and is configured to transfer the control software, flashing program, firmware image file, and flashing tools required for firmware flashing to the Master Linux system running on the ARM MCU chip.

[0076] In a preferred embodiment of the present invention, the GPIO pin of the ARM MCU chip outputs a high-level signal or a low-level signal to the relay switch of the smart PDU so that the relay switch controls the on / off state of the power jack based on the received level signal.

[0077] In a preferred embodiment of the present invention, the Master Linux system also runs a RAMOS server, which integrates DHCP service and NFS service. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing service.

[0078] In a preferred embodiment of the present invention, the control software is configured to determine whether the device node needs to be powered off based on the status information written by each device node. When it is determined that the device node needs to be powered off, the corresponding level signal is sent to the relay switch through the GPIO pin of the ARM MCU chip.

[0079] In a preferred embodiment of the present invention, the flashing program, firmware image file and flashing tool required for flashing the firmware are all stored in a shared folder provided by the NFS service.

[0080] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.

[0081] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.

[0082] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0083] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A system for cluster firmware upgrade, characterized in that, include: The intelligent PDU has a relay switch and multiple power sockets; An ARM MCU chip is disposed within the intelligent PDU, and the GPIO pins of the ARM MCU chip are connected to the relay switch; An RJ45 network interface is connected to the ARM MCU chip, and the RJ45 network interface is configured to connect to a switch and then to each device node in the cluster via the switch; A memory card connected to the ARM MCU chip, a Master Linux system pre-installed in the memory card, and the memory card configured to start the Master Linux system and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on; A USB interface is connected to the ARM MCU chip, and the USB interface is configured to transfer the control software, flashing program, firmware image file and flashing tool required for flashing firmware to the Master Linux system running in the ARM MCU chip; The device node's power supply line is plugged into the power socket of the intelligent PDU, which provides the necessary power for the device node's operation. The device node's network interface is connected to the RJ45 network interface, and the DHCP service in the intelligent PDU provides IP addresses to all devices. After the intelligent PDU is powered on, the Master Linux system in the intelligent PDU starts, the relay switch closes, the device node is powered on, the DHCP service assigns IPMI IPs to the device node, and controls the device node to power on via IPMI commands. The PXE bootloader runs the RAMOS system on the node device, which is mounted on the NFS shared file system. The flashing program captures the local firmware version upgrade list and determines whether the upgrade is complete. If so, the process ends; otherwise, it flashes the firmware for each device in a predefined order. After each flash, it checks whether a power outage or restart is required. If not, it continues to the next device. If a restart is required, the flashing program issues a restart command to the local device to perform the restart. After restarting, it returns to the step of PXE booting the device node to run the RAMOS system and continues to flash other firmware according to the above steps. If a restart is required, the local device status is written to the NFS directory according to the IP identifier. The Linux system obtains information from each node. The Master Linux system issues a disconnect command to the relay. After a set delay, the Master Linux system issues a close command to the relay, which closes. Then, it returns to the step of PXE booting the node device to run the RAMOS system and continues to flash other firmware according to the above steps until all firmware is flashed.

2. The system according to claim 1, characterized in that, The GPIO pins of the ARM MCU chip output high-level or low-level signals to the relay switch of the smart PDU, so that the relay switch controls the on / off state of the power jack based on the received level signal.

3. The system according to claim 1, characterized in that, The Master Linux system also runs a RAMOS server, which integrates DHCP and NFS services. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing services.

4. The system according to claim 1, characterized in that, The control software is configured to determine whether to power off the device nodes based on the status information written by each device node. When it is determined that power off the device nodes is required, the software sends a corresponding level signal to the relay switch through the GPIO pin of the ARM MCU chip.

5. The system according to claim 3, characterized in that, The flashing program, firmware image file, and flashing tools required for firmware flashing are all stored in a shared folder provided by the NFS service.

6. An intelligent supercomputing cluster, characterized in that, The intelligent supercomputing cluster includes a cluster firmware upgrade system, which includes: The intelligent PDU has a relay switch and multiple power sockets; An ARM MCU chip is disposed within the intelligent PDU, and the GPIO pins of the ARM MCU chip are connected to the relay switch; An RJ45 network interface is connected to the ARM MCU chip, and the RJ45 network interface is configured to connect to a switch and then to each device node in the cluster via the switch; A memory card connected to the ARM MCU chip, a Master Linux system pre-installed in the memory card, and the memory card configured to start the Master Linux system and run the Master Linux system on the ARM MCU chip after the smart PDU is powered on; A USB interface is connected to the ARM MCU chip, and the USB interface is configured to transfer the control software, flashing program, firmware image file and flashing tool required for flashing firmware to the Master Linux system running in the ARM MCU chip; The device node's power supply line is plugged into the power socket of the intelligent PDU, which provides the necessary power for the device node's operation. The device node's network interface is connected to the RJ45 network interface, and the DHCP service in the intelligent PDU provides IP addresses to all devices. After the intelligent PDU is powered on, the Master Linux system in the intelligent PDU starts, the relay switch closes, the device node is powered on, the DHCP service assigns IPMI IPs to the device node, and controls the device node to power on via IPMI commands. The PXE bootloader runs the RAMOS system on the node device, which is mounted on the NFS shared file system. The flashing program captures the local firmware version upgrade list and determines whether the upgrade is complete. If so, the process ends; otherwise, it flashes the firmware for each device in a predefined order. After each flash, it checks whether a power outage or restart is required. If not, it continues to the next device. If a restart is required, the flashing program issues a restart command to the local device to perform the restart. After restarting, it returns to the step of PXE booting the device node to run the RAMOS system and continues to flash other firmware according to the above steps. If a restart is required, the local device status is written to the NFS directory according to the IP identifier. The Linux system obtains information from each node. The Master Linux system issues a disconnect command to the relay. After a set delay, the Master Linux system issues a close command to the relay, which closes. Then, it returns to the step of PXE booting the node device to run the RAMOS system and continues to flash other firmware according to the above steps until all firmware is flashed.

7. The cluster according to claim 6, characterized in that, The GPIO pins of the ARM MCU chip output high-level or low-level signals to the relay switch of the smart PDU, so that the relay switch controls the on / off state of the power jack based on the received level signal.

8. The cluster according to claim 6, characterized in that, The Master Linux system also runs a RAMOS server, which integrates DHCP and NFS services. The DHCP service provides IP addresses for the network interfaces on each device node, and the NFS service provides file sharing services.

9. The cluster according to claim 6, characterized in that, The control software is configured to determine whether to power off the device nodes based on the status information written by each device node. When it is determined that power off the device nodes is required, the software sends a corresponding level signal to the relay switch through the GPIO pin of the ARM MCU chip.

10. The cluster according to claim 8, characterized in that, The flashing program, firmware image file, and flashing tools required for firmware flashing are all stored in a shared folder provided by the NFS service.

Images