Large model all-in-one machine dynamic IP system and method

Abstract

The invention relates to the technical field of large-model all-in-one machine deployment, in particular to a large-model all-in-one machine dynamic IP system and method, and the method comprises the following steps: preparing a switch, electrifying the switch, configuring a local area network, starting DHCP support, and setting an address range; putting the deployment server on a rack and accessing the switch, and enabling the switch to automatically start the deployment service after being powered on; a large model all-in-one machine to be filled and deployed is erected and connected into a switch, and PXE starting is set after power-on and power-on; the large-model all-in-one machine has the beneficial effects that high-performance hardware, customized software and large models such as DeepSeek and Qwen are integrated through the large-model all-in-one machine, and one-stop large-model service is provided for enterprises. According to the method, the IP change does not need to be modified by manual intervention when the large-model server is relocated, and the large-model server can be used after being inserted into a network. The service and the physical IP address are decoupled, and the all-in-one machine can be used after being opened. The problem that the IP change service is unavailable is solved, the deployment efficiency is improved, and the batch deployment cost of deliverymen is saved.

Description

Technical Field

[0001] This invention relates to the field of large-scale integrated machine deployment technology, specifically to a system and method for dynamic IP of a large-scale integrated machine. Background Technology

[0002] Currently, ChatGPT, an AI chatbot program based on the GPT-3.5 large language model, has garnered widespread attention worldwide for its fluent language expression, powerful question-handling capabilities, and massive database. Various industries are feeling the powerful impact of large models. Simultaneously, a research boom in large models has officially begun both domestically and internationally. DeepSeek, in particular, has become incredibly popular due to its technological innovation, cost advantages, open-source ecosystem, market demand resonance, and symbolic significance of a breakthrough in domestic AI, resulting in a continuous increase in the delivery volume of its large model products. Due to security requirements, some government agencies prohibit delivery personnel from entering their premises to install large model all-in-one machines, requiring that the machines be ready to use immediately upon delivery, meaning they should be usable right out of the box.

[0003] To enable large-scale integrated machines to be used out of the box and immediately connected to the network and powered on-site, this invention provides a method that supports dynamic IP addresses, allowing large-scale integrated machines to operate normally under different IP addresses in different environments, thus decoupling the physical network card IP address. Summary of the Invention

[0004] The purpose of this invention is to provide a system and method for dynamic IP of large-scale integrated machines, which simplifies the deployment difficulty of large-scale integrated machines, lowers the deployment threshold, enables out-of-the-box use after IP change, and reduces the delivery cost of large-scale integrated machine projects, thereby solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a system for dynamic IP of a large-scale integrated machine, comprising a switch, a deployment server, and a large-scale integrated machine; The switch is used to build a local area network, enable DHCP support and configure address range to provide network access for large model all-in-one machines and deployment servers; After the deployment server is installed and connected to the switch, the deployment service will automatically start upon power-on. This deployment service provides PXE service, detects DHCP network segment IP addresses, and deploys virtual network interface card service upon detection. The virtual network interface card service creates a tap0 virtual network interface card and configures a virtual IP address. The virtual network interface card tap0 is used to deploy a Kubernetes cluster. After successful cluster deployment, the large model service and intelligent agent service components are deployed via Helmchart. After deployment is completed, a deployment detection step is performed. If there are no problems with the deployment, the filling personnel are notified to remove the server. If there are deployment anomalies, an alarm is triggered and the filling personnel are notified to manually handle the anomaly. After the large model all-in-one machine is installed on the rack, it is connected to the switch, powered on and configured to boot via PXE. The operating system is then installed via PXE, and a temporary IP address is obtained via DHCP after the operating system is installed.

[0006] Preferably, in the local area network configuration, the switch is responsible for data exchange and forwarding, ensuring normal network communication between the deployment server and the large-scale integrated machine. Its enabled DHCP supports dynamically allocating IP addresses for access devices, and the configured address range meets the access requirements of devices within the local area network.

[0007] Preferably, after detecting a DHCP network segment IP address, the deployment server starts a virtual network interface service. The tap0 virtual network interface created by this virtual network interface service is used to provide a specific network interface when deploying a k8s cluster. The configured virtual IP address serves as the network identifier for the deployment of the k8s cluster and related services, ensuring the stable deployment and operation of the k8s cluster and related services in a dynamic IP environment.

[0008] Preferably, when the large model all-in-one machine is used on the project site, after connecting the network cable and powering on, the virtual network card service creates a virtual network card tap0 again and sets the VIP address. The physical network card obtains the project site IP address through DHCP. After the k8s service detects the tap0 network card, it starts the k8s cluster. After the k8s cluster starts normally, it launches the large model and intelligent agent services. The customer calls the large model service according to the physical IP of the site.

[0009] Preferably, the system enables the large model appliance to achieve stable operation of network configuration based on dynamic IP, Kubernetes cluster deployment, and large model and agent services in different scenarios through the collaborative work between switches, deployment servers, and large model appliances. This improves the flexibility and efficiency of system deployment and reduces the management and maintenance costs caused by fixed IP configuration.

[0010] A method for dynamic IP of a large-scale all-in-one machine includes the following steps: Prepare the switch, put the switch on the rack and power it on, configure the local area network, enable DHCP support and set the address range; Install the deployment server on the rack and connect it to the switch. Power it on and make it automatically start the deployment service. Mount the large-scale integrated machine to be filled and connected to the switch. After powering it on, set it to PXE boot. The deployment service provides PXE service. The large-scale all-in-one machine boots up via PXE to complete the operating system installation. After installation, it obtains a temporary IP address via DHCP. The deployment service detects DHCP network segment IP addresses. Once detected, it deploys a virtual network interface service, creates a tap0 virtual network interface and configures a virtual IP address. It then uses this virtual network interface to deploy a Kubernetes cluster. After successful cluster deployment, it deploys the large model service and agent service components via Helmchart. After deployment, it performs a test. If there are no problems, it notifies the filling personnel to remove the product from the shelves. If there are abnormalities, it issues an alarm and notifies manual intervention.

[0011] Preferably, when configuring a local area network, the network topology should be planned reasonably according to the actual network size and number of devices to ensure that the switch can exchange data efficiently and stably; the address range set by the enabled DHCP support should be able to meet the IP address allocation needs of all devices in the local area network to avoid address conflicts.

[0012] Preferably, during the deployment of PXE services, it is necessary to ensure that the PXE service is configured correctly and establishes an effective communication connection with the large-scale integrated machine, so that the large-scale integrated machine can successfully boot through PXE and obtain the image file required for operating system installation; after the large-scale integrated machine obtains a temporary IP address, it uses this address to perform normal data interaction with the deployment service for subsequent virtual network card service deployment operations.

[0013] Preferably, after the deployment service detects the DHCP network segment IP address, when the virtual network interface card service creates the tap0 virtual network interface card and configures the VIP address, it ensures that the VIP address is unique within the local area network and does not conflict with the existing device IP address; during the deployment of the Kubernetes cluster using the tap0 virtual network interface card, the deployment specifications and requirements of the Kubernetes cluster are strictly followed to ensure the stability and reliability of the cluster; when deploying large model services and intelligent agent service components via Helmchart, it is ensured that the chart file is configured correctly and that the components can be installed and run normally.

[0014] Preferably, after the model all-in-one machine is filled and transported to the project site, the virtual network card service recreates the tap0 virtual network card and sets the VIP address during startup. During the process of the physical network card obtaining the project site IP address via DHCP, it is ensured that the virtual network card service and the physical network card can work normally and that the obtained IP address is valid. After the k8s service detects the tap0 network card, it starts the k8s cluster. After the cluster starts normally, it launches the large model and intelligent agent services. When the customer calls the large model service according to the physical IP of the site, it ensures the stability and efficiency of the entire service call process and ensures that the customer can use the large model service normally.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes a system and method for dynamic IP in a large-scale model all-in-one appliance. By integrating high-performance hardware, customized software, and large models such as DeepSeek and Qwen into the large-scale model all-in-one appliance, it provides enterprises with a one-stop large-scale model service. It eliminates the need for manual IP modification when relocating the large-scale model server, enabling plug-and-play large-scale model services. It decouples the service from the physical IP address, making the all-in-one appliance ready to use out of the box. It solves the problem of service unavailability due to IP changes, improves deployment efficiency, and saves on batch deployment costs for delivery personnel. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: This invention provides a technical solution: a system for dynamic IP of a large-scale all-in-one machine, comprising: (1) GPU server GPU servers are physical servers used to deploy large-scale models and intelligent agent applications. Depending on the requirements, they include GPUs from different manufacturers and models, such as NVIDIA, Huawei Ascend NPU, Tianshu Zhixin TianGai GPU, and MuXi GPU. After installing the operating system, the GPU server provides the basic physical hardware and software environment for deploying large-scale models and intelligent agent services.

[0019] (2) Network switch Network switches or routers provide temporary IP addresses for GPU servers when loading large model services, which are then used to deploy large models and related services. After the GPU server is installed, it is connected to the network switch. Once powered on, the server installs the operating system and automatically obtains an IP address from the switch's DHCP service.

[0020] (3) K8S cluster Kubernetes, or K8s for short, is an open-source container orchestration engine developed by Google, dedicated to making the deployment of containerized applications simple and efficient. Large model services and intelligent agents are deployed to the Kubernetes cluster using Helm chart templates, achieving service installation, orchestration, and scheduling through containerized deployment, and enabling automatic startup at boot.

[0021] (4) Deployment Service In the filling room, the switches and deployment servers are configured with fixed IP addresses, and the DHCP address range for the switches is set. After the large-scale integrated machine filling server is mounted and connected to the network, PXE is configured to start and power on. The deployment service provides PXE services and installs the operating system on each filling server. After the operating system installation is complete, it automatically obtains a temporary IP address from the switch's DHCP service. The deployment service detects IP addresses in the DHCP network segment. Upon detection, it logs into the GPU server and sequentially deploys components such as the virtual network card service, Kubernetes cluster, large-scale model service, and intelligent agent service to complete the batch filling. After filling is complete, the server is shut down, and the filling personnel are notified to remove the machine and transport it to the project site.

[0022] (5) Virtual network card service When the large-scale integrated appliance is installed in the data center, the GPU server can deploy the Kubernetes cluster and related services after obtaining a temporary TemIP. However, when the server is moved to the customer's site, the server's IP address changes, causing the Kubernetes cluster to fail to start after power-on, thus rendering the large-scale appliance service unavailable. Some government agency customers cannot obtain the project site IP plan in advance, requiring decoupling of the large-scale integrated appliance service startup from the physical network card IP. Even if the customer's IP address plan is available, configuring hundreds or thousands of IP network addresses for installation is required when dozens or hundreds of integrated appliances for different projects are delivered, which is time-consuming and labor-intensive.

[0023] During the large-scale integrated appliance installation process, a virtual network interface card (NIC) service is added, managed via systemd to enable service startup. After the NIC service starts, virtual NICs (such as TUN / TAP or Bridge) are created on the server, and virtual VIP addresses and routes are configured. Then, when deploying the Kubernetes cluster, the virtual NIC and virtual VIP address are used to deploy the cluster, decoupling it from the physical NIC IP addresses. This way, after the large-scale integrated appliances are installed in batches and transported to different customer sites, they only need to be connected to the customer's network. Once powered on, the large-scale service runs normally, using the newly acquired physical IP address at the customer's site.

[0024] Example 2, based on Example 1, proposes a method for dynamic IP of a large-scale all-in-one machine, including the following steps: 1. Switch preparation: Power on the switch, configure the local area network, enable DHCP support, and configure the address range.

[0025] 2. Deploy the server, connect it to the switch, and the deployment service will start automatically after power-on.

[0026] 3. Place the large-scale integrated machine to be filled on the rack, connect it to the switch, power it on, and set it to start with PXE.

[0027] 4. Deployment service provides PXE service, and the large model all-in-one machine boots and installs the operating system through PXE.

[0028] 5. The operating system for the large model all-in-one machine is installed. A temporary IP address is obtained via DHCP.

[0029] 6. The deployment service detects DHCP network segment IP addresses. Once detected, the deployment service deploys the virtual network card service. The virtual network card service creates the tap0 virtual network card and configures the virtual IP address, i.e., configures the VIP address.

[0030] 7. Deploy the service using the virtual network interface card tap0 to deploy the Kubernetes cluster. After the cluster is successfully deployed, deploy components such as the large model service and the agent service using Helm charts.

[0031] 8. After deployment, the deployment service will perform deployment testing steps. If the deployment is successful, the filling personnel will be notified to remove the product from shelves. If a deployment anomaly occurs, an alarm will be triggered, and the filling personnel will be notified to manually handle the anomaly.

[0032] 9. After the large-scale model all-in-one machine is filled and removed from the rack, it is transported to the project site. At the project site, a network cable is connected and the machine is powered on. Upon power-on, the virtual network interface service creates a virtual network interface (tap0) and sets a VIP address. The physical network interface obtains the project site's IP address via DHCP. The Kubernetes service detects the tap0 network interface and starts the Kubernetes cluster. After the Kubernetes cluster starts normally, it launches services such as the large-scale model and intelligent agents. The customer calls the large-scale model service based on the physical IP address at the site.

[0033] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for dynamic IP of a large-scale all-in-one model machine, characterized in that: This includes switches, deployment servers, and large-scale all-in-one machines; The switch is used to build a local area network, enable DHCP support and configure address range to provide network access for large model all-in-one machines and deployment servers; After the deployment server is installed and connected to the switch, the deployment service will automatically start upon power-on. This deployment service provides PXE service, detects DHCP network segment IP addresses, and deploys virtual network interface card service upon detection. The virtual network interface card service creates a tap0 virtual network interface card and configures a virtual IP address. The virtual network interface card tap0 is used to deploy a Kubernetes cluster. After successful cluster deployment, the large model service and intelligent agent service components are deployed via Helmchart. After deployment is completed, a deployment detection step is performed. If there are no problems with the deployment, the filling personnel are notified to remove the server. If there are deployment anomalies, an alarm is triggered and the filling personnel are notified to manually handle the anomaly. After the large model all-in-one machine is installed on the rack, it is connected to the switch, powered on and configured to boot via PXE. The operating system is then installed via PXE, and a temporary IP address is obtained via DHCP after the operating system is installed.

2. The system for dynamic IP of a large-scale integrated machine according to claim 1, characterized in that: In the local area network configuration, the switch is responsible for data exchange and forwarding, ensuring normal network communication between the deployment server and the large-scale integrated machine. Its enabled DHCP supports dynamically assigning IP addresses to access devices, and the configured address range meets the access requirements of devices within the local area network.

3. The system of dynamic IP for a large-scale integrated machine according to claim 2, characterized in that: After detecting a DHCP network segment IP address, the deployment server starts a virtual network interface service. The tap0 virtual network interface created by this service provides a specific network interface when deploying a Kubernetes cluster. The configured virtual IP address serves as the network identifier for the deployment of the Kubernetes cluster and related services, ensuring the stable deployment and operation of the Kubernetes cluster and related services in a dynamic IP environment.

4. The system of dynamic IP for a large-scale integrated machine according to claim 3, characterized in that: When the large model all-in-one machine is used on the project site, after connecting the network cable and powering on, the virtual network card service creates a virtual network card tap0 and sets a VIP address. The physical network card obtains the project site IP address through DHCP. After the k8s service detects the tap0 network card, it starts the k8s cluster. After the k8s cluster starts normally, it launches the large model and intelligent agent services. The customer calls the large model service according to the physical IP address on site.

5. The system for dynamic IP of a large-scale integrated machine according to claim 4, characterized in that: The system enables the large model appliance to achieve stable operation of network configuration based on dynamic IP, Kubernetes cluster deployment, and large model and agent services in different scenarios through the collaborative work of switches, deployment servers, and large model appliances. This improves the flexibility and efficiency of system deployment and reduces the management and maintenance costs caused by fixed IP configuration.

6. A method for dynamic IP of a large-scale all-in-one machine, using the system described in claim 5, characterized in that: Includes the following steps: Prepare the switch, put the switch on the rack and power it on, configure the local area network, enable DHCP support and set the address range; Install the deployment server on the rack and connect it to the switch. Power it on and make it automatically start the deployment service. Mount the large-scale integrated machine to be filled and connected to the switch. After powering it on, set it to PXE boot. The deployment service provides PXE service. The large-scale all-in-one machine boots up via PXE to complete the operating system installation. After installation, it obtains a temporary IP address via DHCP. The deployment service detects DHCP network segment IP addresses. Once detected, it deploys a virtual network interface service, creates a tap0 virtual network interface and configures a virtual IP address. It then uses this virtual network interface to deploy a Kubernetes cluster. After successful cluster deployment, it deploys the large model service and agent service components via Helmchart. After deployment, it performs a test. If there are no problems, it notifies the filling personnel to remove the product from the shelves. If there are abnormalities, it issues an alarm and notifies manual intervention.

7. The method for dynamic IP of a large-scale all-in-one machine according to claim 6, characterized in that: When configuring a local area network (LAN), plan the network topology reasonably according to the actual network size and number of devices to ensure that the switch can exchange data efficiently and stably; the address range set by the enabled DHCP support should be able to meet the IP address allocation needs of all devices in the LAN to avoid address conflicts.

8. The method for dynamic IP of a large-scale all-in-one machine according to claim 7, characterized in that: During the deployment of PXE services, it is necessary to ensure that the PXE service is configured correctly and establishes an effective communication connection with the large-scale integrated machine. This ensures that the large-scale integrated machine can successfully boot through PXE and obtain the image file required for operating system installation. After obtaining a temporary IP address, the large-scale integrated machine uses this address to perform normal data interaction with the deployment service for subsequent virtual network interface card service deployment operations.

9. The method for dynamic IP of a large-scale all-in-one machine according to claim 8, characterized in that: After the deployment service detects the IP address of the DHCP network segment, when the virtual network card service creates the tap0 virtual network card and configures the VIP address, ensure that the VIP address is unique within the local area network and does not conflict with the existing device IP address. When deploying a Kubernetes cluster using the tap0 virtual network interface card, strictly follow the deployment specifications and requirements of the Kubernetes cluster to ensure the stability and reliability of the cluster. When deploying large model services and agent service components using Helmchart, ensure that the chart file is configured correctly and that the components can be installed and run normally.

10. The method for dynamic IP of a large-scale all-in-one machine according to claim 9, characterized in that: After the model all-in-one machine is filled, removed from the shelf, and transported to the project site, the virtual network interface service recreates the tap0 virtual network interface and sets the VIP address during startup. While the physical network interface obtains the project site IP address via DHCP, it ensures that both the virtual network interface service and the physical network interface work normally and that the obtained IP address is valid. After the Kubernetes service detects the tap0 network interface, it starts the Kubernetes cluster. Once the cluster starts normally, it launches the large model and agent services. When the customer calls the large model service based on the on-site physical IP, it ensures the stability and efficiency of the entire service call process, guaranteeing that the customer can use the large model service normally.

Images