Network configuration method and apparatus, related device, storage medium and computer program product

Abstract

The application discloses a network configuration method and device, related equipment, a storage medium and a computer program product, and applies to the technical field of cloud computing. The method comprises the following steps: in response to an application request of a user applying for a bare metal server, network configuration information is respectively allocated to a plurality of neural network processing units (NPUs) in a target bare metal server according to link layer discovery protocol information of the plurality of NPUs, and parameter plane network information of the target bare metal server is obtained; based on the parameter plane network information, a configuration driver file of the target bare metal server is updated; and the target bare metal server is controlled to execute the configuration driver file, and the configuration driver file triggers the target bare metal server to perform network configuration on the plurality of NPUs when being executed.

Description

Technical Field

[0001] This application relates to the field of cloud computing technology, and in particular to a network configuration method, apparatus, related equipment, storage medium and computer program product. Background Technology

[0002] With the continuous development and popularization of cloud computing technology, cloud-based applications have become a trend. In the field of cloud computing technology, Infrastructure as a Service (IaaS) is becoming increasingly mature. Users no longer need to build and maintain physical hardware themselves; they can simply apply for cloud service resources according to their actual needs. In some cases, users require more control, more hardware access, higher processing performance, higher security isolation, and the ability to choose their operating environment. In these situations, a bare metal server (BMS) is a good choice.

[0003] Bare metal servers can provide users with a near-native computing experience, compensating for the significant performance degradation of traditional virtualization instances. In traditional cloud computing and computing networks, bare metal servers serve as a crucial foundational computing resource. Bare metal servers can have independent parameter plane networks, requiring configuration specifically for these networks. However, current technologies for configuring bare metal server parameter plane networks are inefficient; therefore, efficiently configuring these networks is a pressing issue that needs to be addressed. Summary of the Invention

[0004] To address the related technical issues, embodiments of this application provide a network configuration method, apparatus, related devices, storage medium, and program product.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides a network configuration method applied to a cloud management platform, which manages at least multiple bare-metal servers. The method includes:

[0007] In response to a user's request to apply for a bare metal server, network configuration information is allocated to the multiple Neural Processing Units (NPUs) in the target bare metal server according to their Link Layer Discovery Protocol (LLDP) information, thereby obtaining the parameter plane network information of the target bare metal server. The parameter plane network information includes the LLDP information and the network configuration information, and the network configuration information of each NPU corresponds to the LLDP information.

[0008] Based on the parameter plane network information, update the configuration driver file of the target bare metal server;

[0009] The target bare metal server is controlled to execute the configuration driver file, wherein, when the configuration driver file is executed, the target bare metal server triggers the target bare metal server to perform network configuration on the multiple NPUs based on the parameter plane network information.

[0010] In the above scheme, the cloud management platform includes a bare metal deployment component and a network service component. The bare metal deployment component is at least used to request the parameter plane network information of the target bare metal server, and the network service component is at least used to allocate network configuration information to the NPUs of the target bare metal server. The step of allocating network configuration information to the multiple NPUs based on their LLDP information to obtain the parameter plane network information of the target bare metal server includes:

[0011] The bare metal deployment component calls the first interface to send a request message to the network service component. The first interface is used for communication between the bare metal deployment component and the network service component. The request message is used to request the parameter plane network information of the target bare metal server. The request message carries the LLDP information of the multiple NPUs.

[0012] In response to the request message, the network service component allocates network configuration information to the multiple NPUs based on the LLDP information of the multiple NPUs, thereby obtaining the parameter plane network information of the target bare metal server.

[0013] The network service component returns a response message to the bare metal deployment component, wherein the response message carries the parameter plane network information.

[0014] In the above scheme, the configuration driver file includes a user data script, which is used to provide parameter plane network information for the network configuration of the multiple NPUs; updating the configuration driver file of the target bare metal server based on the parameter plane network information includes:

[0015] The parameter plane network information is added to the user data script in the configuration driver file to obtain the updated configuration driver file.

[0016] In the above scheme, the application request carries user data parameters; before updating the configuration driver file of the target bare metal server, the method further includes:

[0017] Script for obtaining user data based on the aforementioned user data parameters;

[0018] The user data script is saved in the configuration driver file.

[0019] In the above scheme, the cloud management platform further includes a computing service component, which is used to transmit the user data parameters carried in the application request to the bare metal deployment component; and to obtain a user data script based on the user data parameters, including:

[0020] The computing service component transmits the user data parameters carried in the application request to the bare metal deployment component;

[0021] The bare metal deployment component obtains the user data script based on the user data parameters.

[0022] In the above scheme, the LLDP information includes one or more of the following: the Media Access Control (MAC) address of the NPU, the port identifier of the uplink switch to which the NPU is connected, the name of the uplink switch, the Internet Protocol (IP) address of the uplink switch management system, and the virtual LAN identifier of the port of the NPU.

[0023] The network configuration information includes one or more of the following: NPU IP address, gateway, and subnet mask.

[0024] In the above scheme, before responding to a user's request for a bare metal server, the method further includes:

[0025] Send a self-test command to the plurality of bare metal servers, the self-test command being used to instruct the plurality of bare metal servers to report LLDP information;

[0026] Receive LLDP information returned by the multiple bare metal servers.

[0027] In the above scheme, when the configuration driver file is executed, the target bare metal server is also triggered to query the network configuration information corresponding to the MAC address in the parameter plane network information using the MAC address of each NPU as an index, and to perform network configuration on the NPU according to the network configuration information.

[0028] In the above scheme, in response to a user's request for a bare metal server, network configuration information is allocated to the multiple NPUs according to the LLDP information of the multiple NPUs in the target bare metal server, thereby obtaining the parameter plane network information of the target bare metal server, including:

[0029] In response to the application request, a target bare metal server is determined from the plurality of bare metal servers;

[0030] If the target bare metal server is configured with multiple NPUs, then network configuration information is allocated to each of the multiple NPUs according to the LLDP information of the multiple NPUs, thereby obtaining the parameter plane network information of the target bare metal server.

[0031] This application also provides a network configuration method applied to a user equipment, wherein the user equipment is at least used to display an access page for requesting a bare metal server; the method includes:

[0032] A request to apply for a bare metal server is sent to the cloud management platform. This request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units (NN units) in the target bare metal server based on their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on this parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information for each NN unit corresponds to the link layer discovery protocol information. When the configuration driver file is executed, it triggers the target bare metal server to configure the network for the multiple NN units based on the parameter plane network information.

[0033] Present the feedback results of the application request.

[0034] This application embodiment also provides a network configuration device, disposed on a cloud management platform, the cloud management platform being used to manage at least multiple bare metal servers, the network configuration device comprising:

[0035] The allocation module is used to respond to a user's request for a bare metal server, allocate network configuration information to the multiple NPUs in the target bare metal server according to the LLDP information of the multiple NPUs, and obtain the parameter plane network information of the target bare metal server. The parameter plane network information includes the LLDP information and the network configuration information, and the network configuration information of each NPU corresponds to the LLDP information. Based on the parameter plane network information, the configuration driver file of the target bare metal server is updated.

[0036] The control module is used to control the target bare metal server to execute the configuration driver file, wherein when the configuration driver file is executed, it triggers the target bare metal server to perform network configuration on the multiple NPUs based on the parameter plane network information.

[0037] This application embodiment also provides a network configuration device, installed in a user equipment, the user equipment being used to display an access page for requesting a bare metal server, the network configuration device comprising:

[0038] The sending module is used to send an application request for a bare metal server to the cloud management platform. The application request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units in the target bare metal server based on the link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on the parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information of each neural network processing unit corresponds to the link layer discovery protocol information. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration for the multiple neural network processing units based on the parameter plane network information.

[0039] The display module is used to present the feedback results of the application request.

[0040] This application also provides a cloud management platform, including: a processor and a memory for storing computer programs that can run on the processor; wherein, when the processor runs the computer programs, it executes the steps of the cloud management platform side method described above.

[0041] This application also provides a user equipment, including: a processor and a memory for storing a computer program that can run on the processor; wherein, when the processor runs the computer program, it executes the steps of the user equipment-side method described above.

[0042] This application embodiment also provides a storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, it implements the steps of the cloud management platform side method described above, or, when the computer program is executed by a processor, it implements the steps of the user equipment side method described above.

[0043] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the cloud management platform-side method described above, or, when executed by a processor, implements the steps of the user equipment-side method described above.

[0044] The network configuration method, apparatus, related devices, storage media, and computer program products provided in this application embodiment can, in response to a user's application request during the bare metal server application process, allocate network configuration information to each NPU in the bare metal server based on the LLDP information of multiple NPUs in the bare metal server. Each NPU's LLDP information corresponds to its network configuration information, thereby enabling the allocation of network configuration information to specific NPUs, thus achieving dynamic allocation of network configuration information for the bare metal server's parameter plane network. Furthermore, the configuration driver file driving the bare metal server is updated using the parameter plane network information of each NPU. This parameter plane network information includes LLDP information and network configuration information. Therefore, during the startup process using the updated configuration driver file, the bare metal server performs parameter plane network configuration for each NPU according to the LLDP information and network configuration information provided by the parameter plane network information. This automatically configures the parameter plane network information for each NPU, reducing manual intervention and achieving efficient configuration of the bare metal server's parameter plane network, saving manpower. Attached Figure Description

[0045] Figure 1 A flowchart illustrating a network configuration method provided in an embodiment of this application;

[0046] Figure 2 A flowchart illustrating an application example of a network configuration method provided in this application embodiment;

[0047] Figure 3 This is a schematic diagram of the structure of a network configuration device provided in an embodiment of this application;

[0048] Figure 4 This is a schematic diagram of another network configuration device provided in an embodiment of this application;

[0049] Figure 5 This is a schematic diagram of the structure of a cloud management platform provided in an embodiment of this application;

[0050] Figure 6This is a schematic diagram of the structure of a user equipment provided in an embodiment of this application. Detailed Implementation

[0051] The present application will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] Bare metal servers typically refer to physical servers or hardware devices without pre-installed operating systems or other software. They combine the performance of virtual machines and physical machines, providing dedicated physical servers in the cloud. They can offer superior computing performance and data security for core databases, critical application systems, high-performance computing, big data, and other applications. Bare metal servers are often simply referred to as bare metal servers.

[0053] Bare metal servers can support various types of processing units, such as a Central Processing Unit (CPU), an NPU, and a Graphics Processing Unit (GPU), among others (one or more can be understood as at least one). A bare metal server can support one or more of each type of processing unit. NPUs are commonly used in artificial intelligence and high-performance computing, especially in deep learning tasks. A bare metal server can be configured with one or more NPUs. For example, a bare metal server can support eight NPUs. For bare metal servers with multiple NPUs, these NPUs can work collaboratively, such as co-processing large-scale distributed artificial intelligence training tasks, improving the performance of the bare metal server. A bare metal server configured with multiple NPUs can be called an intelligent bare metal server.

[0054] In a bare metal service architecture, the cloud management platform serves as the control center. It is responsible for abstracting, pooling, scheduling, and monitoring all bare metal server resources. The cloud management platform includes a dedicated bare metal server management plane. It can be hosted on one or more devices. On the cloud management platform, bare metal servers are represented by bare metal nodes (ironic nodes), enabling software control over the hardware (i.e., the bare metal servers). One bare metal server corresponds to one bare metal node.

[0055] In addition to controlling bare metal servers, the cloud management platform also provides users with an access page to apply for bare metal servers. User devices access the cloud management platform via a browser or client, displaying the application page for bare metal servers. This allows users to order, rent, or try out bare metal servers. For example, the access page provided by the cloud management platform can list multiple customization options for the bare metal server, such as processing unit type, processing unit model, and memory configuration. Users can select these options to apply for, rent, or use the bare metal server. After a successful application, the cloud management platform grants the user access to the bare metal server and returns feedback to the user device, such as a successful application message. This allows users to remotely access and configure the bare metal server through the access page on their devices. This approach satisfies users' needs for personalized customization of bare metal servers.

[0056] Typically, bare metal servers communicate via a Virtual Private Cloud (VPC) plane network. The VPC plane network is a traditional cloud network built upon VPCs. Cloud management platforms usually use the VPC plane network to issue commands, manage logs, and perform other management operations on bare metal servers.

[0057] To meet different data transmission requirements, bare metal servers can also use different networks for communication. For example, for intelligent bare metal servers, in addition to the VPC plane network, a data plane network and a parameter plane network are also added.

[0058] A data plane network is a network channel used for high-speed data transmission. For example, in machine learning scenarios, before or during model training, a bare-metal intelligent computing server can transmit training data through a data plane network.

[0059] Parametric surface networks (PSNs) are network channels used to synchronize parameters between NPUs. For example, in machine learning scenarios, intelligent bare-metal servers can use PSNs to transmit gradient parameters or other model parameters calculated by the NPUs to ensure model consistency. PSNs can be used for high-frequency (e.g., microsecond-level communication cycles), small-size (e.g., as small as a few bytes), globally synchronized data communication.

[0060] The VPC plane network, data plane network, and parameter plane network are physically or logically separated. This makes it difficult for bare-metal server management operations or data loading tasks to preempt the bandwidth and low-latency network resources required for parameter synchronization, thereby improving the stability and performance of AI training tasks.

[0061] For bare metal servers with a parameter plane network, in order to enable communication between the bare metal servers and the parameter plane network, it is necessary to configure the parameter plane network of the bare metal servers.

[0062] One approach to configuring the parameter plane network of a bare metal server is the static configuration scheme. The basic idea of ​​static configuration is that, before the bare metal server is ready, such as during the bare metal service architecture setup, administrators plan the network configuration information for the parameter plane network for each bare metal server. Specifically, under the control of administrators, the cloud management platform statically configures the network configuration information of the bare metal server's parameter plane network into the network configuration file `network-config` of the corresponding bare metal node, thus pre-saving the corresponding bare metal server's parameter plane network configuration information in the bare metal node. After the cloud management platform statically configures the parameter plane network into the network configuration file `network-config` of the bare metal node, administrators can control the cloud management platform to set the bare metal server to an application-ready state. During the bare metal server application and activation process, the cloud management platform initializes the bare metal server using the cloud initialization tool `cloud-init`. During initialization, the cloud management platform drives the bare metal server to configure the network configuration information of the parameter plane network in the network configuration file `network-config` onto the NPU. This achieves static configuration of the bare metal server's parameter plane network.

[0063] Understandably, network configuration information can include IP address, subnet mask, gateway, and other information. Bare-metal servers can configure the NPU's parameter plane network using this information. After successful parameter plane network configuration, high-speed data transmission and synchronization can be achieved between different NPUs.

[0064] Understandably, before a bare metal server is ready, it is in an unavailable state, and users cannot apply for it through the access page provided by the cloud management platform. Conversely, once a bare metal server is available, users can apply for it through the access page provided by the cloud management platform.

[0065] For the static configuration scheme of the aforementioned parameter plane network, the parameter plane network information needs to be manually planned for each bare metal server during the bare metal server construction phase. Before the bare metal servers are ready, the parameter plane network information also needs to be manually configured into the bare metal nodes set in the cloud management platform. This parameter plane network configuration method consumes excessive manpower. In addition, if the network configuration information of the parameter plane network is incorrect, it may cause a chain reaction of problems, resulting in low overall deployment efficiency. For example, for the upcoming Wanka cluster with thousands of bare metal servers, including more than 2,000 intelligent computing bare metal servers, if an address conflict occurs, it will affect the use of parameter planes on multiple intelligent computing bare metal servers in the cluster, and locating and modifying the problem is very troublesome, time-consuming, and labor-intensive. Furthermore, in the current deployment environment, the network configuration information of the parameter plane network is transferred from the cloud management platform to the bare metal server by the administrator using an Excel spreadsheet containing all NPU network configuration information. This requires the administrator to be very familiar with the overall NPU configuration process, and to manually record and manage the network configuration information of all intelligent computing bare metal parameter plane networks. Moreover, the work needs to be interrupted if the file is lost, which is a significant risk.

[0066] Based on this, various embodiments of this application provide a dynamic configuration scheme for the parameter plane network of a bare metal server. When the bare metal server is in an available state, in response to a user's request to apply for a bare metal server, the cloud management platform allocates network configuration information to each NPU in the bare metal server based on the LLDP information of multiple NPUs, thereby realizing the dynamic allocation of network configuration information for the parameter plane network of the bare metal server. Further, the cloud management platform combines the LLDP information and network configuration information of each NPU to obtain the parameter plane network information of each NPU, and uses this parameter plane network information to update the configuration driver file of the bare metal server. Thus, during the startup process using the updated configuration driver file, the bare metal server can configure the parameter plane network of each NPU according to the LLDP information and network configuration information provided by the parameter plane network information. This enables efficient configuration of the bare metal server parameter plane network, improves a series of problems caused by manual configuration of the bare metal server parameter plane network, reduces the complexity of bare metal server operation and maintenance, saves human resources, improves the reliability of the bare metal resource pool, reduces the operation and maintenance burden, and is suitable for promotion to other bare metal clusters.

[0067] It is understandable that in the parameter plane network information generated by the cloud management platform, the LLDP information of each NPU corresponds to the network configuration information of the corresponding NPU. Therefore, when the cloud management platform drives the bare metal server to configure the network configuration information of the parameter plane network in the NPU, the network configuration information of the parameter plane network is configured in the NPU indicated by the LLDP information, realizing a one-to-one matching configuration between the NPU and the network configuration information.

[0068] The parameter plane network dynamic configuration scheme provided in this application embodiment automatically obtains the parameter plane network configuration information during the bare metal service application and activation process, and dynamically configures the parameter plane network configuration information onto the NPU of the bare metal server. This saves manual planning and configuration of network configuration information before the bare metal server is ready, improving the configuration efficiency of the parameter plane network of the bare metal server. Furthermore, the network configuration information provided by the cloud management platform corresponds one-to-one with the NPU, thereby improving the accuracy of the parameter plane network configuration.

[0069] The network configuration method provided in this application embodiment is applied to a cloud management platform, such as... Figure 1 As shown, the method includes the following steps:

[0070] Step 101: The cloud management platform receives the user's application request for a bare metal server;

[0071] Step 102: The cloud management platform responds to the application request and allocates network configuration information to the multiple NPUs in the target bare metal server according to the LLDP information of the multiple NPUs in the target bare metal server, thereby obtaining the parameter plane network information of the target bare metal server;

[0072] Step 103: The cloud management platform updates the configuration driver file of the target bare metal server based on the parameter plane network information of the target bare metal server;

[0073] Step 104: The cloud management platform controls the target bare metal server to execute the configuration driver file.

[0074] In practical application, in step 101, the user equipment (UE) receives the user's request to apply for a bare metal server. In response to this user action, the UE sends a request to the cloud management platform to apply for the bare metal server. The cloud management platform receives the request from the UE. For example, the UE can display an access page for the bare metal server to the user. This access page includes a confirmation button for the bare metal server application. The UE receives the user's action of clicking the confirmation button. In response to this action, the UE sends the request to the cloud management platform to apply for the bare metal server.

[0075] The user equipment described in this application embodiment can be a mobile phone, tablet computer, desktop computer, laptop computer, handheld computer, notebook computer, super mobile personal computer, netbook, as well as cellular phone, personal digital assistant, augmented reality / virtual reality device, media player, wearable device, etc. This application embodiment does not impose any special limitations on the specific form of the user equipment.

[0076] In some optional implementations, the request sent by the user device to the cloud management platform may include user data parameters, such as user-data parameters. For example, user data parameters may include bare metal server specifications, username, password, and one or more parameters from the user's initial network. The bare metal server specifications may include server model, CPU model and quantity, GPU model and quantity, NPU model and quantity, memory configuration, etc.

[0077] In some optional implementations, the bare metal server access page offers various bare metal server options. Users can select the appropriate bare metal server configuration based on their actual needs. Different bare metal server configurations correspond to different bare metal server options. For example, the access page may offer Bare Metal Server Option 1 and Bare Metal Server Option 2. Bare Metal Server Option 1 corresponds to a bare metal server configured with one CPU and one GPU. The bare metal server corresponding to Bare Metal Server Option 2 is configured with one CPU and eight NPUs.

[0078] In some alternative implementations, the bare metal server's access page can also configure customization options. These options provide the bare metal server's configuration requirements. Different combinations of customization options can create bare metal servers of varying specifications. For example, the access page might offer customization options for the number of GPUs and the number of NPUs. If a user selects both the 1-GPU and 8-GPU customization options, the customized bare metal server will be configured with 1 GPU and 8 NPUs.

[0079] In practical application, in step 102, the cloud management platform manages multiple bare metal servers. Responding to a user's request for a bare metal server, the cloud management platform identifies a target bare metal server from among these servers. The target bare metal server is the one among the multiple bare metal servers managed by the cloud management platform that meets the user's customized conditions. These user-customized conditions indicate the specifications of the bare metal server requested by the user. For example, if the user-customized conditions indicate that the requested bare metal server is a bare metal server with 1 CPU and 2 NPUs, then the target bare metal server has 1 CPU and 2 NPUs. The target bare metal server may be configured with at least one NPU. Of course, the target bare metal server may also not be configured with any NPUs.

[0080] In some optional implementations, the application request can include user-defined conditions, such as the aforementioned user data parameters. The cloud management platform uses the bare metal server specifications provided by the user data parameters in the application request to determine the target bare metal server with the corresponding bare metal server specifications.

[0081] In other alternative implementations, the cloud management platform responds to the request by obtaining user-customized conditions from the user's device, such as user-customized conditions including the aforementioned user data parameters. Based on the bare-metal server specifications provided by the user data parameters, the cloud management platform determines the target bare-metal server with the corresponding bare-metal server specifications.

[0082] The LLDP information of multiple NPUs in the target bare metal server can be pre-stored by the cloud management platform, or the LLDP information of multiple NPUs in the target bare metal server can be obtained by the cloud management platform from the target bare metal server after the target bare metal server is identified.

[0083] In some optional implementations, before providing bare metal server request services to users, such as when the cloud management platform establishes communication connections with multiple bare metal servers in the bare metal service architecture, it sends self-test commands to multiple bare metal servers. These self-test commands instruct the multiple bare metal servers to report LLDP information. For example, the cloud management platform executes a self-test command, such as the `inspect` command, instructing the multiple bare metal servers to return the NPU's LLDP information. The bare metal servers respond to the self-test commands by returning the NPU's LLDP information to the cloud management platform. Through these self-test commands, the cloud management platform can automatically collect relevant information from the bare metal servers, thus providing a data foundation for the intelligent allocation of subsequent network configuration information.

[0084] For example, the LLDP information of an NPU includes at least one of the following: the NPU's MAC address, the port identifier of the uplink switch to which the NPU is connected, the name of the uplink switch to which the NPU is connected, the IP address of the uplink switch to which the NPU is connected, and the port-based VLAN identifier (PVID) of the NPU port. The LLDP information of an NPU is hardware information related to the NPU, and a specific NPU can be quickly located using this information.

[0085] The MAC address of an NPU serves as its unique hardware identifier. Different NPUs have different MAC addresses. For example, the MAC address of an NPU can be represented as "14:46:58:7d:ed:d1".

[0086] The port identifier of the uplink switch to which the NPU is connected is the number or name of the port on the uplink switch that establishes the physical connection between the NPU and the uplink switch. For example, the uplink switch port identifier can be represented as "400GE1 / 1 / 1:1". The uplink switch port identifier to which the NPU is connected can be used to locate the physical port on which the NPU is connected to the uplink switch.

[0087] The management IP address of the uplink switch to which the NPU is connected is used to indicate the network location of the uplink switch. For example, the management IP address of the uplink switch to which the NPU is connected can be represented as "125.15.2.201".

[0088] The PVID of an NPU is the default Virtual Local Area Network (VLAN) number set on the network port of the NPU on the VPC side, and it is usually an integer between 1 and 4094. For example, the PVID of an NPU can be represented as "600".

[0089] Understandably, in response to self-test commands, bare metal servers may return self-test information to the cloud management platform in addition to LLDP information. For example, besides LLDP information, bare metal servers can return self-test information such as network configuration, storage configuration, and hardware configuration to the cloud management platform. Hardware configuration may include information such as processing unit type, number of processing units, and processing unit model. Of course, some bare metal servers may not have NPUs configured; in this case, the bare metal server can indicate the absence of NPUs in the hardware configuration information returned to the cloud management platform.

[0090] In practical applications, some bare metal servers may not have NPUs configured. Therefore, in some optional implementations, before the cloud management platform allocates network configuration information to multiple NPUs of the target bare metal server, it determines whether the target bare metal server has NPUs configured. If the target bare metal server has multiple NPUs configured, the cloud management platform allocates network configuration information to each NPU based on their LLDP information. If the target bare metal server does not have multiple NPUs configured, it indicates that the target bare metal server does not require parameter plane network configuration, and the cloud management platform directly controls the startup of the target bare metal server, enabling the target bare metal server for the user.

[0091] For example, the aforementioned network configuration information includes at least one of the NPU's IP address, gateway, and subnet mask. The cloud management platform allocates network configuration information for the parameter plane network to each NPU based on the LLDP information of each NPU in the target bare metal server. Taking one NPU as an example, the cloud management platform determines the IP resource pool corresponding to the parameter plane network of that NPU based on its LLDP information. Further, the cloud management platform selects an available IP address from this IP resource pool and allocates it to the NPU.

[0092] In this embodiment, after the cloud management platform allocates network configuration information for the parameter plane network to each NPU in the target bare metal server, the cloud management platform binds the network configuration information and LLDP information of each NPU, such as associating the allocated IP address with the MAC address of the NPU, to obtain the parameter plane network information of the target bare metal server. The parameter plane network information includes LLDP information and network configuration information of multiple NPUs. There is a corresponding relationship between the network configuration information and LLDP information of each NPU, so that the network configuration information allocated to that NPU can be found through the LLDP information of that NPU, achieving consistency and stability of the NPU parameter plane network configuration.

[0093] For example, the parameter plane network information may include multiple items from the following: the NPU's MAC address, the port identifier of the uplink switch to which the NPU is connected, the name of the uplink switch to which the NPU is connected, the management IP address of the uplink switch to which the NPU is connected, the NPU's PVID, IP address, gateway, and subnet mask. The parameter plane network information of the target bare metal server can be represented by an information list.

[0094] For example, the LLDP information of an NPU includes: uplink switch management IP address "11.12.201.1", MAC address "0a:1b:2c:3d:4e:5f", and uplink switch port identifier "100GE 1 / 0 / 32". The cloud management platform assigns network configuration information to this NPU including: IP address "10.100.0.2", gateway "10.100.0.1", and subnet mask "24". After the cloud management platform assigns network configuration information to this NPU, the resulting parameter plane network information is: uplink switch management IP address "11.12.201.1", MAC address "0a:1b:2c:3d:4e:5f", uplink switch port identifier "100GE 1 / 0 / 32", IP address "10.100.0.2", gateway "10.100.0.1", and subnet mask "24".

[0095] It is understandable that the above example only lists the parameter plane network information for one NPU. In the case of a target bare-metal server comprising n NPUs, the parameter plane network information list will include parameter plane network information for all n NPUs.

[0096] In some optional implementations, the cloud management platform includes a bare metal deployment component and a network service component. The bare metal deployment component is the management component within the cloud management platform. The network service component is the component responsible for network configuration within the cloud management platform. The cloud management platform can send a request message to the network service component via the bare metal deployment component's first interface. This request message requests the parameter plane network information of the target bare metal server. The request message carries LLDP information for multiple NPUs. Responding to the request message, the network service component allocates network configuration information to each NPU based on the LLDP information, thus obtaining the parameter plane network information of the target bare metal server.

[0097] The first interface mentioned above is used for communication between the bare metal deployment component and the network service component. For example, consider an ironic bare metal deployment component, a neutron network service component, and a neutronclient interface as the first interface. Ironic is a service component in the cloud management platform used for automated management of bare metal servers. Neutron is a component in the cloud management platform used to provide network services. The neutronclient interface is the interface for communication between ironic and neutron in the cloud management platform. In the example, ironic calls the neutronclient interface to send a request message to neutron. Neutron responds to ironic, allocating network configuration information for multiple NPUs and returning a response message to ironic, which carries parameter plane network information. Through the bare metal deployment component and the network service component, the cloud management platform can allocate network configuration information to NPUs based on LLDP information, providing automated network service support for NPU parameter plane network configuration.

[0098] Understandably, the request message sent by the bare metal deployment component to the network service component may carry other information besides the LLDP information of multiple NPUs. For example, in addition to the LLDP information of multiple NPUs, the bare metal deployment component may also carry the user port identifier in the request message. The user port identifier is used to indicate the port where the user-customized bare metal server is located. The response message returned by the network service component to the bare metal deployment component may also carry other information besides the parameter plane network information of multiple NPUs. For example, the network service component may also carry the user port identifier, bare metal server identifier, and bare metal server IP address in the response message.

[0099] In practical applications, the configuration driver file for a bare metal server is used to start the server. The cloud management platform maintains the configuration driver files for each bare metal server. When a bare metal server starts up, the configuration driver file provides the necessary drivers, configuration files, etc., for the server to boot.

[0100] After obtaining the parameter plane network information of the target bare metal server, the cloud management platform can write this information into the target bare metal server's configuration driver file (such as configdrive), thus updating the configuration driver file. The updated configuration driver file records the parameter plane network information for multiple NPUs. When the configuration driver file is executed, it triggers the target bare metal server to configure the network for multiple NPUs based on the parameter plane network information. Therefore, during the startup process of the target bare metal server, when the configuration driver file is executed, the parameter plane network information in the configuration driver file can be used to configure the parameter plane network for multiple NPUs.

[0101] In some optional implementations, the cloud management platform can add the target bare metal server's parameter plane network information to the user data script (such as a user-data script) in the target bare metal server's configuration driver file, thereby updating the target bare metal server's configuration driver file. This user data script records initialization commands, user data parameters, environment configuration parameters, etc., and is a user-specific customized script. Adding the target bare metal server's parameter plane network information to the user data script can automatically configure the NPU parameter plane network during the target bare metal server's startup phase.

[0102] As can be understood, as mentioned above, in response to a user's request, the cloud management platform receives user data parameters, such as bare metal server specifications, username, password, and one or more parameters from the user's initial network. After obtaining these user data parameters, the cloud management platform retrieves a user data script. This user data script records user-related information, thus it is a customized configuration script designed to meet the user's specific needs for the bare metal server.

[0103] For example, a cloud management platform deploys a compute service component and a bare metal deployment component. Both the compute service component and the bare metal deployment component are used for the management of bare metal nodes. The compute service component is also used to pass user data parameters to other compute service components. Taking Nova as the compute service component and Ironic as the bare metal deployment component as an example, the cloud management platform calls Ironic through Nova, passing user data parameters to Ironic. After receiving the user data parameters passed by Nova, Ironic obtains a user data script based on the user data parameters, such as writing the user data parameters into the user data script, or generating a user data script carrying the user data parameters.

[0104] In practical application, in step 104, after updating the configuration driver file of the target bare metal server, the cloud management platform sends the configuration driver file to the target bare metal server, instructing it to start. For example, the cloud management platform sends a Preboot Execution Environment (PXE) startup instruction to the target bare metal server, along with the updated configuration driver file. In this way, the target bare metal server can start via PXE and configure the NPU's parameter plane network through the configuration driver file during the startup process.

[0105] The target bare metal server responds to startup instructions from the cloud management platform, such as a PXE startup instruction, and begins startup. During the PXE process on the target bare metal server, the parameter plane network information of each NPU is obtained from the user data script in the configuration driver file. Based on the NPU's parameter plane network information, the parameter plane network of each NPU is configured. For example, the target bare metal server can look up the MAC address of the corresponding NPU by its serial number. Then, using this MAC address as an index, the network configuration information corresponding to that MAC address is retrieved from the user data script, and the network configuration information corresponding to that MAC address is further configured for that NPU, thus realizing the NPU parameter plane network configuration.

[0106] If the target bare metal server starts successfully, it returns a success message to the cloud management platform. Upon receiving this message, the cloud management platform then sends a response to the user device, indicating a successful application. The user device can then display this successful application result on its access page, allowing remote access to the customized bare metal server.

[0107] If the target bare metal server fails to start, it returns a startup failure response message to the cloud management platform. Alternatively, if the cloud management platform does not receive a response message, it considers the target bare metal server to have failed to start. After receiving this response message, the cloud management platform returns feedback on the application request to the user device, such as a failure response. The user device can display the failure response on the access page, allowing the user to reapply for a bare metal server from the cloud management platform.

[0108] The parameter plane network configuration scheme provided in this application embodiment can realize the dynamic allocation of network configuration information of bare metal servers in the parameter plane network, thereby improving the configuration efficiency of bare metal server parameter plane network and saving human and physical resources.

[0109] The following section provides a more detailed description of this application with reference to application examples.

[0110] In this application example, taking the computing service component of the cloud management platform as Nova, the bare metal deployment component as Ironic, the network service component as Neutron, and the aforementioned first interface as the NeutronClient interface as an example, the network configuration method flow provided in this application embodiment is illustrated exemplarily. Figure 2 As shown, the method includes the following steps:

[0111] Step 201: When building a bare metal service architecture, the cloud management platform sends a self-test command to the bare metal servers it manages;

[0112] The cloud management platform executes the inspect command (an example of a self-test command), instructing multiple bare-metal servers connected via the VPC plane network to perform self-tests and returning self-test information. This self-test information includes one or more of the following: network configuration, storage configuration, hardware configuration, and NPU LLDP information.

[0113] Step 202: The bare metal server responds to the self-test command and returns self-test information to the cloud management platform;

[0114] The bare metal server responds to the inspect command, performs a self-test, obtains the self-test information, and returns the self-test information to the cloud management platform.

[0115] Step 203: The cloud management platform determines whether the bare metal server has successfully completed its self-test;

[0116] For any bare metal server among multiple bare metal servers, if the cloud management platform successfully receives the self-test information returned by that bare metal server, the cloud management platform considers the bare metal server's self-test to be successful and proceeds to step 204. If the cloud management platform does not successfully receive the self-test information returned by that bare metal server, the cloud management platform considers the bare metal server's self-test to have failed and returns to step 201.

[0117] After self-testing, the cloud management platform can save the obtained self-test information in the corresponding bare metal node of the bare metal server. The cloud management platform can provide the self-test information obtained from the bare metal server self-test through the `node-port-list` subcommand of `ironic`.

[0118] For example, the self-detection information of the bare metal service provided by the cloud management platform through node-port-list includes: Universally Unique Identifier (UUID) "e628cfb4-5a6d-4b3e-a61d-864d85d92fbb", MAC address "98:f0:83:d7:26:c6", uplink switch name "JSFH-604H05-P4F1-POD6-ROCE-JR-CE8875-8", uplink switch management IP address "10.150.241.39", uplink switch port identifier "200GE1 / 0 / 10", PVID "2311", application status "False", processing unit type "npu", and NPU number "npu0, npu1, npu2, npu3, npu4, npu5, npu6, npu7, npu8".

[0119] Step 204: The cloud management platform sets the application status of the bare metal service to an "application available" status, and then proceeds to step 205;

[0120] Before the bare metal server successfully completes its self-test, it is in an unavailable application status, and users cannot apply for it through the access page provided by the cloud management platform. After the bare metal server successfully completes its self-test, the cloud management platform changes the application status from unavailable to available, and users can then apply for a bare metal server through the access page provided by the cloud management platform.

[0121] For example, the bare metal nodes maintained by the cloud management platform have status information indicating the application status of bare metal servers. When the status information in the bare metal node is the first type, such as "False," it means that the bare metal server corresponding to that node is not available for application. When the status information in the bare metal node is the second type, such as "True," it means that the bare metal server corresponding to that node is available for application. The cloud management platform sets the application status of the bare metal service to "available" by changing the status information from the first type to the second type.

[0122] Step 205: The cloud management platform responds to the user's application request and determines whether the bare metal server ordered by the user is a smart computing bare metal server;

[0123] Under the user's control, the user device sends an application request to the cloud management platform to order a bare metal server, such as an application request to order a bare metal server with 8 NPUs.

[0124] In response to a user's request, the cloud management platform identifies the bare metal server ordered by the user and initiates the activation process for that server via Nova. During the activation process, the cloud management platform uses Nova to first determine whether the bare metal server has an NPU. For example, Nova can use the self-test information obtained from the bare metal server's self-test to determine whether the bare metal server ordered by the user is configured with an NPU.

[0125] If the bare metal server ordered by the user is configured with an NPU, it means that the bare metal server ordered by the user is a smart computing bare metal server, and the cloud management platform will proceed to step 206.

[0126] If the bare metal server ordered by the user is not configured with an NPU, it means that the bare metal server ordered by the user is not a smart computing bare metal server, and the cloud management platform will proceed to step 210.

[0127] Step 206: The cloud management platform will call ironic through nova, triggering ironic to execute the bare metal service activation process;

[0128] The cloud management platform calls ironic via nova, passing the user data parameters obtained by the cloud management platform when it receives a user's request to order a bare metal server to ironic, thus triggering ironic to execute the bare metal service activation process. The user data parameters can be used to obtain user data scripts.

[0129] Step 207: The cloud management platform calls the neutronclient interface through ironic to request parameter plane network information from neutron;

[0130] The cloud management platform's ironic receives user data parameters from nova, saves these parameters in a user data script, and calls the neutronclient interface to send a request message to neutron. This request message carries parameters required by neutron, including the user port identifier and LLDP information (e.g., represented as NPU_infos). For example, in a bare-metal server configured with 8 NPUs, the request message would include a list of LLDP information for all 8 NPUs and their user port identifiers.

[0131] Step 208: Has the Neutron of the cloud management platform successfully returned the parameter plane network information to the Ironic?

[0132] The cloud management platform uses Neutron to allocate network configuration information to the NPUs of the bare metal server, such as assigning IP addresses, gateways, and subnet masks. If Neutron successfully allocates the network configuration information, it associates the NPU's LLDP information with the network configuration information to obtain the NPU's parameter plane network information and returns a response message to Ironic. This response message carries the parameter plane network information of the eight NPUs, as well as the user port identifier, bare metal server identifier, and bare metal server's IP address. The cloud management platform then proceeds to step 209.

[0133] If neutron fails to allocate network configuration information or fails to return parameter plane network information to ironic, the cloud management platform will execute step 215.

[0134] Step 209: The ironic of the cloud management platform injects the parameter plane network information into the user data script, and then executes step 210;

[0135] Once ironic successfully obtains the parameter plane network information of the bare metal server, it injects this information as a string into the user-data script, separating the parameter plane network information for different NPUs with commas. Furthermore, ironic can return the user-data script containing the parameter plane network information to nova.

[0136] Step 210: The cloud management platform sends a configuration driver file containing user data scripts to the bare metal server to control the bare metal server to start.

[0137] The cloud management platform sends a PXE boot instruction to the bare metal server and includes an updated user data script in the configuration driver file, thereby controlling the bare metal server's startup. The bare metal server responds to the PXE boot instruction and enters the PXE boot process.

[0138] Step 211: The bare metal server determines whether the user image file has been successfully installed;

[0139] During the PXE boot process, the bare metal server installs the user image file provided by the cloud management platform in the configuration driver file to meet the user's required operating environment after boot.

[0140] If the bare metal server successfully installs the user image file, then execute S212.

[0141] If the user image file is not successfully installed on the bare metal server, proceed to step 215.

[0142] Step 212: The bare metal server determines whether the user data script contains parameter plane network information;

[0143] During the startup process of the bare metal server, when executing the host system's cloud initialization (cloud-init), the user data script is automatically run. While running the user data script, the bare metal server determines whether the user data script contains parameter plane network information.

[0144] If the user data script contains parameter plane network information, the bare metal server executes step 213.

[0145] If the user data script does not contain parameter plane network information, the bare metal server will proceed to step 214.

[0146] Step 213: Configure bare metal server parameters for the network plane;

[0147] During the execution of the user-data script, the bare metal server downloads a parameter plane driver installation script, such as auto_install_ecloud_NPU.sh, to adapt the hardware and software. After the parameter plane driver is installed, the bare metal server parses the parameter plane network information from the user-data script. Using a network configuration tool, such as hccn_tool, it sequentially queries the MAC address of the NPU corresponding to its serial number, using the MAC address as the unique matching value. It then finds the corresponding IP address, gateway, subnet mask, and other network configuration information within the parameter plane network information and configures the parameter plane networks of multiple NPUs one by one using the network configuration tool.

[0148] Step 214: Bare metal server startup complete, end current processing flow;

[0149] Once the bare metal server has started up, users can access the corresponding bare metal service and remotely control the bare metal server.

[0150] Step 215: The cloud management platform terminates the activation process and ends the current processing flow.

[0151] If the bare metal server fails to be activated, the cloud management platform will throw an exception message, print an exception log, and terminate the activation process of the bare metal server.

[0152] The solution provided in this application example, after automatically configuring and obtaining parameter plane network information, adds this information to the user data script of the bare metal server's customized activation process. This ensures the parameter plane network information meets user customization needs and improves consistency. Furthermore, using the NPU's MAC address as a matching item enhances the accuracy of the parameter plane network configuration. Compared to static configuration methods requiring manual intervention, this eliminates the need for manual intervention, reducing the probability of errors due to cumbersome network configuration on the bare metal server and lowering the likelihood of server failure. This application example solidifies and automates the network configuration process, reducing labor costs, improving the efficiency of parameter plane network configuration, lowering the barrier to entry for NPU network configuration, and making changes to NPU network configuration intelligent, significantly reducing the operational burden.

[0153] To implement the network configuration method provided in this application embodiment, this application embodiment also provides a network configuration device, applied to a cloud management platform, for managing at least multiple bare metal servers, such as... Figure 3 As shown, the network configuration device includes:

[0154] The allocation module 31 is used to respond to the user's request for a bare metal server, allocate network configuration information to multiple NPUs according to the LLDP information of multiple NPUs in the target bare metal server, and obtain the parameter plane network information of the target bare metal server. The parameter plane network information includes LLDP information and network configuration information, and the network configuration information of each NPU corresponds to the LLDP information. Based on the parameter plane network information, the configuration driver file of the target bare metal server is updated.

[0155] Control module 32 is used to control the target bare metal server to execute the configuration driver file, wherein when the configuration driver file is executed, it triggers the target bare metal server to perform network configuration on multiple NPUs based on parameter plane network information.

[0156] In one possible implementation, the cloud management platform includes a bare metal deployment component and a network service component. The bare metal deployment component is at least used to request the parameter plane network information of the target bare metal server, and the network service component is at least used to allocate network configuration information to the NPU of the target bare metal server. The bare metal deployment component sends a request message to the network service component through the allocation module 31 by calling a first interface. The first interface is used for communication between the bare metal deployment component and the network service component. The request message is used to request the parameter plane network information of the target bare metal server, and the request message carries the LLDP information of the plurality of NPDs.

[0157] The network service component responds to the request message through the allocation module 31, and allocates network configuration information to the multiple NPDs based on the LLDP information of the multiple NPDs, thereby obtaining the parameter plane network information of the target bare metal server.

[0158] The network service component returns a response message to the bare metal deployment component through the allocation module 31, wherein the response message carries the parameter plane network information.

[0159] In one possible implementation, the configuration driver file includes a user data script that provides parameter plane network information for multiple NPDs to configure the network.

[0160] The allocation module 31 adds the parameter plane network information to the user data script of the configuration driver file to obtain the updated configuration driver file.

[0161] In one possible implementation, the allocation module 31 is further configured to:

[0162] Script for obtaining user data based on the aforementioned user data parameters;

[0163] The user data script is saved in the configuration driver file.

[0164] In one possible implementation, the cloud management platform further includes a computing service component; the computing service component transmits the user data parameters carried in the application request to the bare metal deployment component through the allocation module 31, wherein the computing service component is used to transmit the user data parameters carried in the application request to the bare metal deployment component;

[0165] The bare metal deployment component obtains the user data script based on user data parameters through the allocation module 31.

[0166] In one possible implementation, the allocation module 31 can also be used for:

[0167] Send self-test commands to multiple bare metal servers. The self-test commands are used to instruct the multiple bare metal servers to report LLDP information.

[0168] Receive LLDP information returned by multiple bare metal servers.

[0169] In one possible implementation, the allocation module 31 is specifically used for:

[0170] In response to the application request, a target bare metal server is determined from the plurality of bare metal servers;

[0171] If the target bare metal server is configured with multiple NPUs, then network configuration information is allocated to each of the multiple NPUs according to the LLDP information of the multiple NPUs, thereby obtaining the parameter plane network information of the target bare metal server.

[0172] In practical applications, the control module 32 can be implemented by the processor in the network configuration device in conjunction with the communication interface, and the allocation module 31 can be implemented by the processor in the network configuration device.

[0173] This application also provides a network configuration device applied to a user equipment, which is at least used to display an access page for requesting a bare metal server, such as... Figure 4 As shown, the network configuration device includes:

[0174] The sending module 41 is used to send an application request for a bare metal server to the cloud management platform. The application request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units in the target bare metal server based on the link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on the parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information of each neural network processing unit corresponds to the link layer discovery protocol information. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration on the multiple neural network processing units based on the parameter plane network information.

[0175] Display module 42 is used to present the feedback results of the application request.

[0176] In one possible implementation, the network configuration device further includes a display module 42 for displaying an access page for applying for a bare metal server when the bare metal server is in an applyable state.

[0177] It should be noted that the network configuration device provided in this application embodiment is only illustrated by the above-described division of program modules. In actual applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the processing described above. Furthermore, the network configuration device and network configuration method provided in this application embodiment belong to the same concept, and their specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0178] Based on the hardware implementation of the above program modules, and in order to implement the cloud management platform method of this application embodiment, this application embodiment also provides a cloud management platform, such as... Figure 5 As shown, the cloud management platform includes:

[0179] The communication interface 501 is capable of exchanging information with bare metal servers or user equipment;

[0180] The processor 502 is connected to the communication interface 501 to enable information interaction with the bare metal server or user equipment, and to execute the methods provided by one or more technical solutions on the cloud management platform side when running computer programs.

[0181] The computer program is stored in memory 503.

[0182] Specifically, the communication interface 501 is used to receive a request from a user equipment to apply for a bare metal server;

[0183] The processor 502 is used for:

[0184] In response to a user's request to apply for a bare metal server, network configuration information is allocated to the multiple NPUs according to the LLDP information of the multiple NPUs in the target bare metal server, thereby obtaining the parameter plane network information of the target bare metal server. The parameter plane network information includes the LLDP information and the network configuration information, and the network configuration information of each NPU corresponds to the LLDP information.

[0185] Based on the aforementioned parameter plane network information, update the configuration driver file of the target bare metal server; and,

[0186] The target bare metal server is controlled to execute the configuration driver file via communication interface 501. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration on the multiple NPUs based on the parameter plane network information.

[0187] In some optional embodiments, the cloud management platform includes a bare metal deployment component and a network service component. The bare metal deployment component is at least used to request the parameter plane network information of the target bare metal server, and the network service component is at least used to allocate network configuration information to the NPUs of the target bare metal server. The bare metal deployment component sends a request message to the network service component through the processor 502 by calling a first interface, wherein the first interface is used for communication between the bare metal deployment component and the network service component, and the request message is used to request the parameter plane network information of the target bare metal server, and the request message carries the LLDP information of the plurality of NPUs.

[0188] The network service component responds to the request message through the processor 502, and allocates network configuration information to the multiple NPUs based on the LLDP information of the multiple NPUs, thereby obtaining the parameter plane network information of the target bare metal server;

[0189] The network service component returns a response message to the bare metal deployment component through the processor 502, wherein the response message carries the parameter plane network information.

[0190] In some optional embodiments, the configuration driver file includes a user data script, which is used to provide parametric network information for the multiple neural network processing units to configure the network; the processor 502 is specifically used to add the parametric network information to the user data script of the configuration driver file to obtain an updated configuration driver file.

[0191] In some optional embodiments, the cloud management platform further includes a computing service component, which is used to transmit the user data parameters carried in the application request to the bare metal deployment component;

[0192] The computing service component transmits the user data parameters carried in the application request to the bare metal deployment component through the processor 502;

[0193] The bare metal deployment component obtains the user data script based on the user data parameters through the processor 502.

[0194] In some optional embodiments, the communication interface 501 is further used for:

[0195] Send a self-test command to the plurality of bare metal servers, the self-test command being used to instruct the plurality of bare metal servers to report LLDP information;

[0196] Receive LLDP information returned by the multiple bare metal servers.

[0197] In some alternative embodiments, the processor 502 is specifically used for:

[0198] In response to the application request, a target bare metal server is determined from the plurality of bare metal servers;

[0199] If the target bare metal server is configured with multiple NPUs, then network configuration information is allocated to each of the multiple NPUs according to the LLDP information of the multiple NPUs, thereby obtaining the parameter plane network information of the target bare metal server.

[0200] It should be noted that the specific processing procedures of the processor 502 and the communication interface 501 can be understood by referring to the above method.

[0201] Of course, in practical applications, the various components in the cloud management platform are coupled together through bus system 504. It can be understood that bus system 504 is used to implement communication between these components. In addition to the data bus, bus system 504 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 5 The general designated all buses as Bus System 504.

[0202] The memory 503 in this embodiment is used to store various types of data to support the operation of the cloud management platform. Examples of such data include any computer programs used to operate on the cloud management platform.

[0203] The methods disclosed in the embodiments of this application can be applied to the processor 502, or implemented by the processor 502. The processor 502 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 502 or by instructions in the form of software. The processor 502 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 502 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in the memory 403. The processor 502 reads the information in the memory 503 and combines its hardware to complete the steps of the aforementioned method.

[0204] In an exemplary embodiment, the cloud management platform may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned methods.

[0205] Based on the hardware implementation of the above program modules, and in order to implement the user equipment side method of the embodiments of this application, the embodiments of this application also provide a user equipment, such as... Figure 6 As shown, the user equipment includes:

[0206] The communication interface 601 enables information exchange with the cloud management platform;

[0207] The processor 602 is connected to the communication interface 601 to enable information interaction with the cloud management platform and to execute the methods provided by one or more technical solutions on the user equipment side when running computer programs.

[0208] The computer program is stored in the memory 603.

[0209] Specifically, the communication interface 601 is used to send an application request for a bare metal server to the cloud management platform. The application request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units in the target bare metal server based on the link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on the parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information of each neural network processing unit corresponds to the link layer discovery protocol information. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration for the multiple neural network processing units based on the parameter plane network information.

[0210] The processor 602 displays the feedback result of the application request on the screen.

[0211] It should be noted that the specific processing procedures of the communication interface 601 and the processor 602 can be understood by referring to the user equipment side method described above, and will not be repeated here.

[0212] Of course, in practical applications, the various components in the user equipment are coupled together through the bus system 604. It can be understood that the bus system 604 is used to implement communication between these components. In addition to the data bus, the bus system 604 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 6 The general designated all buses as Bus System 604.

[0213] The memory 603 in this embodiment is used to store various types of data to support the operation of the user device. Examples of such data include any computer program used to operate on the user device.

[0214] The methods disclosed in the embodiments of this application can be applied to or implemented by processor 602. Processor 602 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 602 or by instructions in the form of software. Processor 602 may be a general-purpose processor, DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 602 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 603. Processor 602 reads the information in memory 603 and combines it with its hardware to complete the steps of the aforementioned method.

[0215] It is understood that the memory 503 or memory 603 in the embodiments of this application can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable types of memories.

[0216] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 503 storing a computer program, which can be executed by the processor 502 of a cloud management platform to complete the steps described in the aforementioned cloud management platform-side method. Another example is a memory 603 storing a computer program, which can be executed by the processor 602 of a user device to complete the steps described in the aforementioned user device-side method. The computer-readable storage medium can be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0217] In an exemplary embodiment, this application also provides a computer program product, including a computer program that can be executed by a processor 402 of a cloud management platform to complete the steps described in the aforementioned cloud management platform-side method.

[0218] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0219] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0220] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A network configuration method, characterized in that, Applied to a cloud management platform, wherein the cloud management platform is used to manage at least multiple bare metal servers, the method includes: In response to a user's request to apply for a bare metal server, network configuration information is allocated to the multiple neural network processing units in the target bare metal server according to the link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information of each neural network processing unit corresponds to the link layer discovery protocol information. The link layer discovery protocol information includes one or more of the following: the MAC address of the neural network processing unit, the port identifier of the uplink switch to which the neural network processing unit is connected, the name of the uplink switch, the management IP address of the uplink switch, and the virtual LAN identifier of the port of the neural network processing unit. Based on the parameter plane network information, update the configuration driver file of the target bare metal server; The target bare metal server is controlled to execute the configuration driver file, wherein, when the configuration driver file is executed, the target bare metal server triggers the target bare metal server to perform network configuration on the plurality of neural network processing units based on the parameter plane network information.

2. The method according to claim 1, characterized in that, The cloud management platform includes a bare metal deployment component and a network service component. The bare metal deployment component is at least used to request the parameter plane network information of the target bare metal server, and the network service component is at least used to allocate network configuration information to the neural network processing units of the target bare metal server. The step of allocating network configuration information to the multiple neural network processing units in the target bare metal server according to their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server, includes: The bare metal deployment component calls the first interface to send a request message to the network service component. The first interface is used for communication between the bare metal deployment component and the network service component. The request message is used to request the parameter plane network information of the target bare metal server. The request message carries the link layer discovery protocol information of the plurality of neural network processing units. In response to the request message, the network service component allocates network configuration information to the multiple neural network processing units based on the link layer discovery protocol information of the multiple neural network processing units, thereby obtaining the parameter plane network information of the target bare metal server. The network service component returns a response message to the bare metal deployment component, wherein the response message carries the parameter plane network information.

3. The method according to claim 1, characterized in that, The configuration driver file includes a user data script, which provides parameter plane network information for the network configuration of the plurality of neural network processing units; updating the configuration driver file of the target bare metal server based on the parameter plane network information includes: The parameter plane network information is added to the user data script in the configuration driver file to obtain the updated configuration driver file.

4. The method according to claim 3, characterized in that, The application request carries user data parameters; before updating the configuration driver file of the target bare metal server, the method further includes: Script for obtaining user data based on the aforementioned user data parameters; The user data script is saved in the configuration driver file.

5. The method according to any one of claims 1 to 4, characterized in that, The network configuration information includes one or more of the following: the IP address of the neural network processing unit, the gateway, and the subnet mask.

6. The method according to any one of claims 1 to 4, characterized in that, Prior to responding to a user's request for a bare metal server, the method further includes: Send a self-test command to the plurality of bare metal servers, the self-test command being used to instruct the plurality of bare metal servers to report link layer discovery protocol information; Receive link layer discovery protocol information returned by the plurality of bare metal servers.

7. The method according to any one of claims 1 to 4, characterized in that, When the configuration driver file is executed, the target bare metal server is also triggered to query the network configuration information corresponding to the MAC address in the parameter plane network information using the MAC address of each neural network processing unit as an index, and to perform network configuration on the neural network processing unit according to the network configuration information.

8. The method according to any one of claims 1 to 4, characterized in that, In response to a user's request for a bare metal server, network configuration information is allocated to the multiple neural network processing units in the target bare metal server based on their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server, including: In response to the application request, a target bare metal server is determined from the plurality of bare metal servers; If the target bare metal server is configured with multiple neural network processing units, then network configuration information is allocated to each of the multiple neural network processing units according to the link layer discovery protocol information of the multiple neural network processing units, thereby obtaining the parameter plane network information of the target bare metal server.

9. A network configuration method, characterized in that, Applied to user equipment, the user equipment is at least used to display an access page for requesting a bare metal server; the method includes: A request to apply for a bare metal server is sent to the cloud management platform. This request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units (NN units) on the target bare metal server based on their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on this parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information for each NN unit corresponds to the link layer discovery protocol information. The link layer discovery protocol information includes one or more of the following: the MAC address of the NN unit, the port identifier of the uplink switch connected to the NN unit, the name of the uplink switch, the management IP address of the uplink switch, and the virtual LAN identifier of the port of the NN unit. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration for the multiple NN units based on the parameter plane network information. Present the feedback results of the application request.

10. A network configuration device, characterized in that, The network configuration device is set up on a cloud management platform, which is used to manage at least multiple bare metal servers, and includes: The allocation module, in response to a user's request for a bare metal server, allocates network configuration information to the multiple neural network processing units (NN units) within the target bare metal server based on their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information for each NN unit corresponds to the link layer discovery protocol information. The link layer discovery protocol information includes one or more of the following: the MAC address of the NN unit, the port identifier of the uplink switch to which the NN unit is connected, the name of the uplink switch, the management IP address of the uplink switch, and the virtual LAN identifier of the port of the NN unit. Based on the parameter plane network information, the module updates the configuration driver file of the target bare metal server. The control module is used to control the target bare metal server to execute the configuration driver file, wherein, when the configuration driver file is executed, it triggers the target bare metal server to perform network configuration on the multiple neural network processing units based on the parameter plane network information.

11. A network configuration device, characterized in that, The network configuration device is configured in a user equipment, which is at least used to display an access page for requesting a bare metal server. A sending module is used to send a request to the cloud management platform to apply for a bare metal server. The request triggers the cloud management platform to allocate network configuration information to the multiple neural network processing units (NN units) in the target bare metal server based on their link layer discovery protocol information, thereby obtaining the parameter plane network information of the target bare metal server. Based on this parameter plane network information, the platform updates the configuration driver file of the target bare metal server and controls the target bare metal server to execute the configuration driver file. The parameter plane network information includes the link layer discovery protocol information and the network configuration information. The network configuration information for each NN unit corresponds to the link layer discovery protocol information. The link layer discovery protocol information includes one or more of the following: the MAC address of the NN unit, the port identifier of the uplink switch connected to the NN unit, the name of the uplink switch, the management IP address of the uplink switch, and the virtual LAN identifier of the port of the NN unit. When the configuration driver file is executed, it triggers the target bare metal server to perform network configuration for the multiple NN units based on the parameter plane network information. The display module is used to present the feedback results of the application request.

12. A cloud management platform, characterized in that, include: Processor and memory used to store computer programs that can run on the processor; When the processor is used to run a computer program, it executes the steps of the method according to any one of claims 1 to 8.

13. A user equipment, characterized in that, include: Processor and memory used to store computer programs that can run on the processor; When the processor is used to run a computer program, it executes the steps of the method described in claim 9.

14. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8, or when the computer program is executed by a processor, it implements the steps of the method according to claim 9.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8, or when the computer program is executed by a processor, it implements the steps of the method according to claim 9.

Images