Configuration method and device for high performance computing network, electronic device, storage medium, and computer program product

Abstract

The present application discloses a configuration method and device for a high-performance computing network, an electronic device, a storage medium, and a computer program product, which relate to the field of high-performance computing technology, and include: obtaining network configuration information of a high-performance computing network; sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information, thereby solving the technical problem that traditional high-performance computing network configuration methods rely on manual operations, resulting in low configuration efficiency, and achieving the technical effect of improving configuration efficiency.

Description

Technical Field

[0001] The present application relates to the field of high-performance computing technology, and in particular to a configuration method and device for a high-performance computing network, an electronic device, a storage medium, and a computer program product. Background Art

[0002] With the rapid development of artificial intelligence and high-performance computing (HPC) in recent years, the demand for data exchange between server nodes has increased significantly. This is especially true when training large-scale machine learning models, which places unprecedented demands on network latency and throughput. Traditional Ethernet technology is struggling to meet these demands, especially when dealing with large amounts of data, where the network becomes a performance bottleneck. To address this issue, Remote Direct Memory Access (RDMA) technology has emerged. Its efficient data transmission capabilities significantly reduce latency and improve throughput in data exchange between servers. High-performance computing networks (such as InfiniBand networks), as a specific implementation of RDMA technology, are ideal for connecting servers, storage systems, and network devices in HPC and AI applications due to their extremely low latency and high data transfer rates.

[0003] However, with the rise of intelligent computing centers, the management and configuration of high-performance computing networks face new challenges. To meet the demands of high-performance computing networks in multi-tenant scenarios, cloud platforms must support automated and isolated network configuration. Traditional high-performance computing network configuration methods rely on manual operations, which is not only inefficient but also makes it difficult to effectively isolate and quickly allocate resources in multi-tenant environments.

[0004] In related technologies, traditional high-performance computing network configuration methods rely on manual operations, resulting in low configuration efficiency. No effective solution has been proposed yet. Summary of the Invention

[0005] The present application provides a configuration method and device for a high-performance computing network, an electronic device, a storage medium, and a computer program product to at least solve the problem in the related art that traditional high-performance computing network configuration methods rely on manual operations, resulting in low configuration efficiency.

[0006] The present application provides a configuration method for a high-performance computing network, comprising: obtaining network configuration information of the high-performance computing network; sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information.

[0007] The present application also provides a configuration device for a high-performance computing network, including: an acquisition module for acquiring network configuration information of the high-performance computing network; a configuration module for sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information.

[0008] The present application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing the steps of any of the above-mentioned methods for configuring a high-performance computing network when executing the computer program.

[0009] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods for configuring a high-performance computing network are implemented.

[0010] The present application also provides a computer program product, including a computer program, which implements the steps of any of the above-mentioned methods for configuring a high-performance computing network when executed by a processor.

[0011] Through this application, the high-performance computing network is configured by sending network configuration information to the subnet manager through the network management module, thereby realizing the automated configuration of the high-performance computing network, improving the configuration efficiency, and solving the problem that the traditional high-performance computing network configuration method relies on manual operation, resulting in low configuration efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0013] Figure 1 This is a hardware structure block diagram of a method for configuring a high-performance computing network according to an embodiment of the present application;

[0014] Figure 2 is a flowchart of a method for configuring a high-performance computing network according to an embodiment of the present application;

[0015] Figure 3 is a schematic diagram of a network management module according to an embodiment of the present application;

[0016] Figure 4 is a working diagram of a network adaptation module according to an embodiment of the present application;

[0017] Figure 5 is a schematic diagram of allocating a partitioned network to a cloud host according to an embodiment of the present application;

[0018] Figure 6 This is a structural block diagram of the configuration of a high-performance computing network according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0021] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] In conjunction with the specific application environment architecture or specific hardware architecture on which the execution of the configuration method of the high-performance computing network depends, the specific application environment architecture or specific hardware architecture is described herein.

[0023] The method embodiments provided in the embodiments of the present application can be executed in a server device or a similar computing device. Taking running on a server device as an example, Figure 1 This is a hardware structure diagram of a configuration method for a high-performance computing network according to an embodiment of the present application. Figure 1 As shown, the server device may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. The server device may also include a transmission device 106 and an input / output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above server device. Figure 1More or fewer components than shown, or with Figure 1 Different configurations shown.

[0024] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the startup method of the operating system in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories can be connected to the server device via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0025] Transmission device 106 is used to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by a communication provider of the server device. In one embodiment, transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0026] The embodiments of the present application provide a configuration method for a high-performance computing network, and the method is described in detail in conjunction with the execution flow of the configuration method for a high-performance computing network.

[0027] The following is an explanation of the professional terms that appear in this application:

[0028] Infiniband: A high-performance network protocol standard used primarily in high-performance computing and artificial intelligence.

[0029] OpenStack: The industry's mainstream open source cloud platform, providing computing, storage, networking and other services;

[0030] RDMA: Remote Direct Memory Access technology;

[0031] SR-IOV: Single Root Virtualization, a hardware device virtualization technology that can split a single physical device into multiple virtualized devices.

[0032] In this embodiment, a method for configuring a high-performance computing network is provided, including but not limited to application to a cloud platform (including but not limited to an OpenStack cloud platform). Figure 2 is a flow chart of a configuration method of a high performance computing network according to an embodiment of the present application, such as Figure 2 As shown, the method includes the following steps S202-S204:

[0033] Step S202, obtaining network configuration information of the high performance computing network;

[0034] Step S204: Sending network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information.

[0035] Through the above steps, the high-performance computing network is configured by sending network configuration information to the subnet manager through the network management module, thereby realizing the automated configuration of the high-performance computing network, improving the configuration efficiency, and solving the problem that the traditional high-performance computing network configuration method relies on manual operation, resulting in low configuration efficiency.

[0036] In the embodiment provided in step S202 , a network administrator or a cloud platform user may input network configuration information in an interactive interface of the cloud platform.

[0037] In the embodiment provided in step S204, the network configuration information can be sent to the network management module by calling the API interface of the network management module, and then the network configuration information is sent to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network.

[0038] It should be noted that the subnet manager is OpenSM (Open Subnet Manage).

[0039] It should be noted that after receiving the network configuration information, the network management module will convert the network configuration information into OpenSM configuration through its own configuration management submodule and send it to OpenSM.

[0040] It should be noted that after receiving network configuration information, the subnet manager will perform specific configuration operations based on this information, such as setting up network partitions, allocating ports, and adjusting QoS parameters. During operation, the subnet manager should also have monitoring capabilities to detect any anomalies or errors in the configuration process and promptly provide feedback to the network management module for correction or troubleshooting.

[0041] Optionally, after completing the network configuration, the subnet manager will feed back the configuration results to the network management module, including configuration status, execution results, error information, etc. The network management module can further confirm the configuration results and perform secondary configuration or adjustments to ensure that the network configuration fully meets expectations.

[0042] The automated configuration method in the above steps not only improves configuration efficiency and reduces errors caused by manual intervention, but also makes network configuration more flexible and intelligent, and can quickly respond to the personalized needs of multiple tenants in the cloud platform and changes in the network environment, thereby significantly improving the management capabilities and performance levels of high-performance computing networks.

[0043] As an optional implementation, the above-mentioned step S202 includes: obtaining partition configuration information, and / or network topology information, and / or port configuration information, and / or routing configuration information, and / or service quality configuration information, and / or security configuration information of the high-performance computing network; wherein the network configuration information includes at least one of the following: partition configuration information, network topology information, port configuration information, routing configuration information, service quality configuration information, and security configuration information.

[0044] Optionally, partition configuration information: In high-performance computing networks, partition configuration information is used to define logical segments of the network. Partition configuration information is particularly useful in InfiniBand networks, where network subnets are defined using PKeys and GUIDs (globally unique identifiers) to ensure network resource isolation between different tenants. Obtaining partition configuration information can include determining the number of subnets to be partitioned, the PKey range for each subnet, and the devices and services associated with the subnets.

[0045] Network topology information: This describes the connectivity and structure of devices within the network, including the connections between servers, switches, and storage devices, the network's hierarchical structure, and the location and type of network devices. For high-performance computing networks, topology information is crucial for optimizing data flow paths, balancing network loads, and planning network expansion.

[0046] Port configuration information: Port configuration information involves the settings of the ports on the network device, such as port speed, duplex mode, flow control, error recovery capability, etc. In InfiniBand networks, it is also necessary to configure the port PKey association to implement port access control for network partitions.

[0047] Routing configuration information: Routing configuration information defines the transmission path of data packets in the network, including routing table settings, priority assignments, and failover strategies. In the field of high-performance computing, optimized routing strategies can reduce network latency and improve data transmission efficiency.

[0048] Quality of Service (QoS) configuration information: QoS configuration information allows network administrators to set data transmission priorities, bandwidth limits, and latency sensitivity to ensure that mission-critical network resource requirements are met. In multi-tenant scenarios, QoS configuration also ensures fair allocation of network resources among different tenants.

[0049] Security configuration information: This information includes network access control, encryption settings, firewall rules, and other information to protect the network from unauthorized access and potential attacks. Security configuration is particularly important in high-performance computing networks to ensure the secure transmission of sensitive data and access control to storage resources.

[0050] This configuration information will be used in subsequent automated configuration processes to ensure that the network can intelligently adjust according to pre-defined policies and requirements to meet the various needs of high-performance computing scenarios, such as AI model training and HPC task execution. By automatically acquiring and applying this configuration information, network administrators can significantly reduce the burden of manual configuration, improve the accuracy and efficiency of network configuration, and enhance the network's flexibility and responsiveness to adapt to dynamic changes in high-performance computing environments.

[0051] In an exemplary embodiment, before the above step S204, the method further includes the following steps S11-S12:

[0052] Step S11: converting the acquired configuration information into configuration information of a target format through the network adapter module, wherein the target format is a format corresponding to the application program interface of the network management module, and the configuration information carries the network configuration information;

[0053] Step S12: calling the application program interface of the network management module and sending the configuration information in the target format to the network management module.

[0054] It should be noted that, in this embodiment, OpenSM is configured through an API interface rather than a configuration file, thereby achieving automatic configuration of the high-performance computing network.

[0055] In an exemplary embodiment, when a high-performance computing network has N partitions, each of the N partitions corresponds to a network service in a cloud platform, and the network management module is located in the cloud platform.

[0056] It should be noted that if Figure 4 As shown, this application develops a cloud platform network interface adapter module to receive cloud platform requests, convert them into Infiniband network management module interface format data, call the Infiniband network management module interface, and realize the configuration management of the Infiniband network.

[0057] Taking the OpenStack cloud platform as an example, the cloud platform network adapter module must comply with the OpenStack Neutron ML2 specification to implement OpenStack network, subnet, and port business logic. For example, regarding network isolation, Ethernet uses virtual local area networks (VLANs) to isolate subnets, with different subnets having different VLAN IDs. InfiniBand uses partitions, with partitions isolated using private keys. Therefore, each partition must correspond to an OpenStack network concept.

[0058] It's important to note that when a high-performance computing network uses N partitions, each partition is mapped as an independent network service within the cloud platform, enabling logical isolation and on-demand resource allocation. The network management module, a core component of the cloud platform, centrally manages these services, dynamically configuring network parameters such as bandwidth, latency, and security policies based on user needs to ensure efficient and secure network operation in a multi-tenant environment. This architecture improves resource utilization, simplifies network management, and enhances the flexibility and responsiveness of the cloud platform.

[0059] In an exemplary embodiment, in the process of sending network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, the method further includes: synchronizing the network configuration information to a target database through the network management module.

[0060] like Figure 3 As shown, the network configuration information can be synchronized to the target database through the data synchronization submodule in the network management module.

[0061] In the configuration and management process of a high-performance computing network, the network management module not only sends network configuration information to the subnet manager, but is also responsible for synchronizing this configuration information to the target database. This synchronization mechanism ensures the persistent storage of configuration data and the consistency of network status. Even after a system restart or failure, the network configuration can be quickly restored, avoiding the complexity of manual configuration. Database synchronization also supports auditing and backtracking, making it easier for administrators to track the history of configuration changes, which is particularly important for troubleshooting and compliance reviews in multi-tenant environments. In addition, this mechanism reduces human errors and improves the accuracy and efficiency of network configuration through automated synchronization operations. It is the key to building a stable, efficient, and secure high-performance computing network infrastructure.

[0062] In an exemplary embodiment, the above-mentioned sending of network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network includes: sending a target configuration file to the subnet manager through the network management module, wherein the target configuration file carries the network configuration information; or sending the network configuration information to the subnet manager based on a secure remote access protocol through the network management module.

[0063] It should be noted that the network management module can connect to the subnet manager via a secure remote access protocol. Optionally, the secure remote access protocol is SSH (Secure Shell Protocol).

[0064] It should be noted that when the network management module updates the HPC network configuration, sending network configuration information to the subnet manager (such as OpenSM) is a key step in achieving network status updates. This process can be performed in two ways: one is to send the target configuration file, and the other is to directly send the configuration information based on a secure remote access protocol.

[0065] With the target configuration file sending method, the network management module creates a file containing the latest network configuration information, typically in XML or JSON format. This file is then transferred to the subnet manager node using a file transfer protocol (such as SCP). The subnet manager reads and parses the file, applying the configuration to update the network status. This method is suitable for scenarios with complex and frequently changing configuration information, ensuring configuration consistency and integrity.

[0066] Using secure remote access protocols (such as SSH), the network management module connects directly to the subnet manager over the network, using SSH to encrypt and transmit network configuration information. The subnet manager then immediately updates the configuration upon receiving this information. This approach is particularly suitable for scenarios where real-time configuration information is required or where configuration information is frequently updated, enabling rapid response and improving network configuration efficiency and security.

[0067] Through these two methods, the network management module can ensure that network configuration information is accurately and timely conveyed to the subnet manager, realizing dynamic management and optimization of high-performance computing networks.

[0068] In an exemplary embodiment, the method further includes: synchronizing the network configuration data in the subnet manager to the target database via the network management module at every preset time interval.

[0069] In an exemplary embodiment, the method further includes: when a data synchronization instruction is obtained, synchronizing the network configuration data in the subnet manager to the target database through the network management module.

[0070] Optional, such as Figure 3 As shown, in order to maintain data consistency between the Infiniband network management module and OpenSM, the data between the two can be synchronized regularly and manually through the data synchronization submodule in the network management module.

[0071] Optionally, to ensure consistency and persistence of HPC network configuration data between the cloud platform and the subnet manager, the network management module implements a periodic and on-demand data synchronization strategy. At preset intervals, such as hourly or daily, the network management module proactively extracts the latest network configuration data from the subnet manager (e.g., OpenSM) and synchronizes it to the target database, enabling real-time backup of configuration data. This periodic synchronization mechanism promptly identifies and corrects potential discrepancies in configuration data, ensuring that the cloud platform can quickly restore the correct network configuration upon system restart or failure recovery, thereby improving system stability and reliability.

[0072] Furthermore, upon receiving specific data synchronization instructions, the network management module will immediately initiate an on-demand synchronization process, rapidly synchronizing the subnet manager's network configuration data to the target database, whether following a configuration change or manually triggered by operations personnel. This mechanism enhances network management flexibility, enabling operations personnel to synchronize configuration data instantly based on actual needs, effectively responding to sudden network status changes or urgent troubleshooting needs, and improving network management and maintenance efficiency.

[0073] By combining regular and on-demand data synchronization strategies, the network management module can not only maintain the real-time and consistency of network configuration data, but also provide a flexible response mechanism to ensure the stable operation and efficient management of high-performance computing networks in multi-tenant environments.

[0074] In an exemplary embodiment, the method further includes determining a network status of the high performance computing network based on the network configuration data in the target database.

[0075] It should be noted that since the data synchronization submodule supports timed and manual synchronization mechanisms, it achieves data consistency synchronization between the OpenSM configuration file and the target database. When an OpenSM or underlying device failure occurs, it can be automatically synchronized to the method database and then synchronized to the management platform, improving the overall availability of the method.

[0076] It's important to note that determining the network status of a high-performance computing network is crucial for ensuring the proper allocation of cloud platform resources, monitoring network health and security, and optimizing network performance. By analyzing network configuration data stored in the target database, the network management module accurately depicts the current network topology, the configuration parameters of each subnet (partition), and the association status between cloud hosts and Infiniband network cards, providing a comprehensive understanding of the network's real-time status.

[0077] Specifically, the network management module reads network configuration data from the target database, including but not limited to the port configuration of each Infiniband switch, PKey partition information, SR-IOV virtual function (VF) network card allocation, and the connection status between the cloud host and the Infiniband network. By parsing and processing this data, the network management module generates a network status report that includes the current status of each network component, the validity of the configuration information, and any possible configuration conflicts or anomalies.

[0078] Furthermore, the determination of the network status helps the network management module perform the following functions:

[0079] 1. Resource allocation and optimization: Based on the current network status, network resources are rationally planned and allocated, such as bandwidth, latency, and other key performance indicators, to ensure network service quality for each tenant.

[0080] 2. Fault Detection and Recovery: When anomalies are detected in the network status report, the problem can be quickly located and measures can be taken to recover from the fault or reallocate resources, thereby reducing the impact of network interruptions on cloud platform services.

[0081] 3. Security policy implementation: Adjust security policies, such as access control lists (ACLs), based on network status to ensure the security of network communications and prevent unauthorized access or malicious attacks.

[0082] 4. Performance monitoring and tuning: By continuously monitoring network status and collecting network performance data such as throughput, packet loss rate, and end-to-end latency, we can optimize network performance and improve overall network efficiency.

[0083] 5. User service support: Based on accurate network status, we provide users with customized network service support, such as dynamically adjusting network configuration to meet the computing task requirements of specific tenants, thereby improving user satisfaction.

[0084] Through the network configuration data in the target database, the network management module can accurately grasp the status of the high-performance computing network, provide dynamic, secure, and high-performance network management services for the cloud platform, and meet the complex and changing computing needs in multi-tenant scenarios.

[0085] In an exemplary embodiment, the method further includes the following steps S21-S22:

[0086] Step S21: When a target user applies to create a target cloud host in a target server, a partition of the high-performance computing network selected by the target user is determined, and a target partition identifier is obtained;

[0087] Step S22: The target cloud host is scheduled through the network configuration module so that the target cloud host runs in the network environment corresponding to the target partition identifier.

[0088] It's important to note that when a user requests to create a cloud host on a target server on a cloud platform and explicitly requests the use of a high-performance computing network, such as an InfiniBand network, the user's selected network partition must first be determined. A network partition is a key concept in InfiniBand networks used to isolate network traffic between different tenants and is identified by a specific PKey. The user's partition selection is received through the user interface or API request. The network configuration module then schedules and configures the cloud host based on this target partition identification.

[0089] The above steps not only improve the automation level of cloud host creation, but also optimize resource allocation efficiency, ensuring that each created cloud host can obtain the required InfiniBand network resources, thereby supporting demanding computing tasks such as artificial intelligence training and inference. At the same time, it reduces operation and maintenance complexity and costs, and improves the overall service quality and user experience of the cloud platform.

[0090] In an exemplary embodiment, the target cloud host is scheduled through the network configuration module, including: scheduling a virtual network card for the target cloud host through the network configuration module; associating the network card information of the virtual network card to the partition corresponding to the target partition identifier through the network configuration module, so that the target cloud host runs in the network environment corresponding to the target partition identifier.

[0091] It should be noted that after the target user applies to create a cloud host and specifies a specific partition of the high-performance computing network, the network configuration module first schedules a virtual network card (VF) based on SR-IOV technology for the cloud host within the cloud platform. This VF network card is virtualized from the physical network card (PF) of the high-performance computing network and has independent network resources, including MAC addresses, IP addresses, and InfiniBand-specific port GUIDs and node GUIDs.

[0092] Next, the network configuration module associates the scheduled VF network card information, including its Port GUID and Node GUID, with the user-selected target partition identifier. This process enables granular control of cloud host network access rights, ensuring network isolation between different user cloud hosts and meeting the strict control requirements for network resources in multi-tenant scenarios.

[0093] In an exemplary embodiment, scheduling a virtual network card for a target cloud host through a network configuration module includes: determining a virtual network card in an unmounted state from a virtual network card resource pool; and mounting the virtual network card to the target cloud host.

[0094] It's important to note that to ensure the target cloud host can access and use the HPC network's virtual network interface card (NIC) resources, the network configuration module executes a series of automated scheduling processes. First, the network configuration module examines the virtual network interface card (VF) resource pool, which contains multiple virtual function (VF) NICs generated by virtualizing physical network interfaces (PFs) using SR-IOV technology. Each VF has independent network resources, such as a MAC address, IP address, port GUID, and node GUID, and is initially unattached to any cloud host. The network configuration module screens the resource pool for unattached VFs to ensure that the selected VF is not occupied by other cloud hosts, thereby avoiding network resource conflicts. This process may involve evaluating performance metrics such as availability, network bandwidth, and latency of the VF NICs in the resource pool to ensure optimal network access performance for the target cloud host.

[0095] After selecting the appropriate VF NIC, the network configuration module mounts it to the target cloud host. This mounting process is typically implemented using the cloud platform's virtualization technology. For example, in an OpenStack environment, the module coordinates with the Nova compute service and Neutron network service to mount the selected VF NIC to the target cloud host's virtual machine instance using PCIe passthrough or SR-IOV technology. This ensures that the target cloud host can directly access and utilize the VF NIC's network resources without requiring an additional network stack or virtualization layer, resulting in efficient and low-latency network communication.

[0096] Through the above steps, the network configuration module not only provides high-performance network access capabilities for cloud hosts, but also optimizes the resource allocation process, improves resource utilization, reduces operation and maintenance complexity, and provides strong network support for cloud services in multi-tenant environments.

[0097] In an exemplary embodiment, associating the network card information of the virtual network card with the partition corresponding to the target partition identifier through the network configuration module includes the following steps S31-S32:

[0098] Step S31: Determine the port globally unique identifier corresponding to the virtual network card;

[0099] It should be noted that the Port GUID (Port Globally Unique Identifier) ​​and Node GUID (Node Globally Unique Identifier) ​​corresponding to the virtual network card are pre-configured.

[0100] Step S32: The network configuration module schedules the network management module so as to associate the port globally unique identifier corresponding to the virtual network card with the partition corresponding to the target partition identifier through the network management module.

[0101] In an exemplary embodiment, the above step S32 includes: sending a configuration instruction to the subnet manager through the network management module, so that the subnet manager associates the port globally unique identifier corresponding to the virtual network card with the partition corresponding to the target partition identifier.

[0102] It's important to note that ensuring that virtual network interfaces (VFs) are correctly associated with user-specified network partitions in high-performance computing networks is key to achieving multi-tenant isolation and accurate allocation of network resources. The network configuration module plays a central role in the automated configuration process, working in conjunction with the network management module to achieve this goal.

[0103] When a VF NIC is assigned to a target cloud host, the network configuration module first determines the VF NIC's port globally unique identifier (Port GUID). The Port GUID uniquely identifies a device port in an InfiniBand network and is crucial for associating the VF NIC with a specific network partition.

[0104] The network configuration module then dispatches the network management module, which is responsible for communicating with the subnet manager (such as OpenSM) to perform specific configuration updates. Specifically, the network management module sends configuration instructions to the subnet manager, including the VF NIC's port GUID and target partition identifier (such as the PKey), requesting the subnet manager to update its configuration and set the selected VF NIC's port list and PKey in the target partition. This operation ensures that the VF NIC can only communicate within the designated network partition, achieving precise allocation of network resources and isolation of multi-tenant networks.

[0105] To implement this configuration update, the network management module and the subnet manager may communicate using the SSH protocol to ensure the secure transmission of configuration instructions. After receiving the instructions, the subnet manager will update the configuration of the network partition, save the updated configuration information, and may trigger the recalculation and optimization of the network state to adapt to the new network resource allocation.

[0106] This series of automated processes not only reduces manual intervention and improves configuration efficiency, but also ensures that the high-performance computing network on the cloud platform can accurately respond to user needs, providing customized, securely isolated network environments for different tenants, supporting them to efficiently and securely perform tasks such as artificial intelligence training and reasoning. At the same time, it reduces operation and maintenance costs and improves the overall service quality and user experience of the cloud platform.

[0107] In an exemplary embodiment, the target cloud host is scheduled through the network configuration module so that the target cloud host runs in the network environment corresponding to the target partition identifier. The method also includes: when the target user requests to modify the network configuration of the network environment corresponding to the target partition identifier, determining the modification configuration information; sending the target instruction to the subnet manager through the network management module, so that the subnet manager modifies the network configuration of the partition corresponding to the target partition identifier according to the modification configuration information.

[0108] Optionally, after the target cloud host has successfully run in a specific network partition, the network management module also provides the ability to dynamically modify the network configuration to respond to user changes to the network environment. When the target user requests to adjust the configuration parameters of their network partition, such as changing the PKey or port list, the network management module first captures the modification request, parses it, and determines the specific configuration information to be modified.

[0109] The network management module then sends a target command, including the configuration modification details, to the subnet manager (such as OpenSM). Upon receiving the command, the subnet manager automatically updates the network configuration of the target partition to ensure it aligns with the user's latest requirements. This dynamic configuration capability significantly enhances the flexibility of cloud platform network services, allowing users to adjust the network in real time based on actual business needs, optimizing network performance to meet diverse computing and communication scenarios, while ensuring efficient utilization of network resources and secure isolation in multi-tenant environments.

[0110] Through this mechanism, users can flexibly adjust network partition parameters, such as increasing or decreasing the number of ports, changing PKey settings, and even re-partitioning the network when necessary, without downtime or service interruption. The close collaboration between the Network Management Module and the Subnet Manager enables automated and intelligent network configuration, providing users with an efficient and stable network management and optimization tool, and enhancing the cloud platform's competitiveness in the fields of high-performance computing and artificial intelligence.

[0111] In an exemplary embodiment, the above-mentioned step S204 includes: verifying whether each server node of the high-performance computing network has a high-performance network driver installed and whether the target server node has a subnet manager; when each server node of the high-performance computing network has a high-performance network driver installed and the target server node has a subnet manager, sending network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network.

[0112] It's important to note that when updating the network configuration of a high-performance computing network (such as InfiniBand), the network management module first performs a comprehensive check of the network environment to ensure that all server nodes have correctly installed a high-performance network driver (such as OFED). This is a prerequisite for implementing InfiniBand network functionality. The network management module also verifies that a subnet manager (such as OpenSM) is deployed on the target server nodes, as the subnet manager is responsible for distributing and managing network configurations.

[0113] Once the server node's driver installation status and the existence of a subnet manager are confirmed, the network management module sends network configuration information to the subnet manager via an API interface rather than a traditional configuration file. This information may include specific parameters such as partition configuration updates, port list adjustments, or PKey settings. Upon receiving the configuration instructions, the subnet manager automatically updates its stored network configuration and may trigger a recalculation of the network status to ensure optimal network performance and accurate resource allocation.

[0114] This verification and configuration delivery process ensures automated network configuration, avoiding the complexity and error risks of manual configuration while improving configuration efficiency and network responsiveness. Working closely with the subnet manager, the network management module becomes the core of high-performance computing network configuration, supporting the dynamic allocation and management of network resources in multi-tenant cloud platform scenarios, providing users with a flexible, efficient, and secure network environment.

[0115] In an exemplary embodiment, in a multi-tenant scenario, different tenants often have different business needs, especially in areas such as artificial intelligence training and high-performance computing, which have extremely high requirements for network bandwidth and latency. This application proposes to introduce an intelligent algorithm to dynamically adjust the bandwidth and latency of the Infiniband network to meet the real-time needs of tenants. Specifically:

[0116] 1. Demand perception layer: The cloud platform's monitoring system collects tenants' network usage data in real time, including traffic, latency, and packet loss rate. Based on this data, a network demand model is built for each tenant.

[0117] 2. Intelligent Algorithm Module: This machine learning-based intelligent algorithm module predicts future demand changes based on the tenant's network demand model and the real-time status of network resources, dynamically adjusts network resource allocation, and optimizes bandwidth and latency.

[0118] 3. Resource reservation mechanism: Network resources are reserved for high-priority tenants to ensure stable network performance even when the network is busy.

[0119] 4. Real-time adjustment of network parameters: The network management module will adjust network parameters such as MTU size and transmission protocol in real time based on the output of the algorithm module to reduce latency and improve transmission efficiency.

[0120] In an exemplary embodiment, in order to further enhance the effect of InfiniBand network virtualization, enhanced management of SR-IOV virtualization technology can be performed to achieve more refined virtual function (VF) network card resource allocation and performance optimization. Specifically:

[0121] 1. VF resource pool optimization: Build a VF resource pool to integrate the VF resources generated by virtualization of all physical Infiniband network cards. The resource pool will record the VF usage status, performance parameters and associated cloud host information, and use an adaptive algorithm for resource scheduling.

[0122] 2. VF performance monitoring and optimization: Use the VF performance monitoring tool to monitor the network performance of each VF in real time, such as throughput, latency, and error rate. Based on the monitoring data, the VF configuration parameters, such as MAC address, IP address, Port GUID, and Node GUID, are automatically adjusted to optimize network resource usage.

[0123] 3. Automatic recovery of VF failures: A VF failure detection mechanism is designed. When a VF fails or its performance degrades, the faulty VF is automatically removed from the resource pool and new VF resources are scheduled to the affected cloud host to ensure service continuity and stability.

[0124] In an exemplary embodiment, in order to enhance network security, a role-based access control (RBAC) mechanism is introduced to provide more refined control over access to the multi-tenant Infiniband network. Specifically:

[0125] 1. Role definition and permission allocation: Define roles for different types of tenants and administrators, such as super administrator, general administrator, tenant, etc. Each role corresponds to different network operation permissions, such as creating a network, modifying network configuration, monitoring network status, etc.

[0126] 2. Network policy enforcement: The network management module will enforce the corresponding network policy based on the user role and operation request. For example, a regular administrator may not be able to modify the PKey settings of a network partition, while a super administrator can perform all network operations.

[0127] 3. Audit and Logging: Audit logs record all network operations, including who performed what operation, when and where, and the results of the operation, providing traceability for network events and enhancing security audit capabilities.

[0128] In an exemplary embodiment, in order to improve the overall operation and maintenance efficiency of the cloud platform, an Infiniband network self-diagnosis and repair system is designed to automatically detect network failures and performance bottlenecks, and provide repair suggestions or automatic repairs. Specifically:

[0129] 1. Fault Detection: Integrated network fault detection algorithms monitor the health of the InfiniBand network, such as device failures, line problems, and configuration errors, and generate alerts immediately upon detecting a problem.

[0130] 2. Performance Analysis: Analyze network performance data and identify performance bottlenecks, such as network congestion and latency anomalies, to provide data support for network optimization.

[0131] 3. Self-healing mechanism: Design a network self-healing algorithm that can automatically adjust the network configuration when a network failure or performance issue is detected, such as reconfiguring Pkey and optimizing the port list, to restore network performance.

[0132] In an exemplary embodiment, in order to ensure the stability and traceability of network configuration, a version control and rollback mechanism is introduced to manage the Infiniband network configuration of the cloud platform. Specifically:

[0133] 1. Version control: Infiniband network configurations are stored in a version control system. Each configuration update creates a new version record, including the update time, updater, and update content, enabling historical traceability of configuration changes.

[0134] 2. Configuration rollback: Provides a configuration rollback function. When a new configuration causes network problems, it can quickly roll back to the previous stable version to ensure the continuity of network services.

[0135] 3. Difference comparison: Develop a configuration difference comparison tool that can compare the configuration differences between two versions, helping operations and maintenance personnel quickly locate configuration issues and perform targeted repairs.

[0136] It's important to note that these solutions, respectively targeting intelligent dynamic network adjustments, enhanced SR-IOV management, role-based network access control, a self-diagnosis and repair system, and configuration versioning and rollback mechanisms, comprehensively enhance the performance, security, and operational efficiency of multi-tenant Infiniband network management on cloud platforms. By implementing these solutions, cloud platforms will be able to better meet the network requirements of high-performance computing and artificial intelligence, providing more stable, secure, and efficient services while reducing operational costs and improving user experience.

[0137] In an exemplary embodiment, in a multi-tenant environment, different users and applications have varying requirements for network quality of service (QoS). Traffic prioritization and scheduling are crucial, especially in high-performance computing and artificial intelligence scenarios. This application proposes an adaptive QoS and traffic engineering solution to achieve intelligent management and optimization of traffic in Infiniband networks. Specifically:

[0138] 1. Traffic classification and marking: In the InfiniBand network management module, DSCP (Differentiated Services Code Point) or a similar marking mechanism is introduced to classify and mark different types of traffic, such as training traffic, inference traffic, and management traffic.

[0139] 2. Intelligent QoS Policy: Develop an AI-based QoS policy engine that dynamically adjusts network resource allocation, such as bandwidth, latency, and packet loss rate, based on network load, traffic type, and user priority, ensuring optimal network service for critical applications.

[0140] 3. Traffic Engineering Optimization: Utilizes TE (Traffic Engineering) technology to intelligently adjust network paths and traffic distribution, avoid network congestion, and improve network resource utilization and application performance.

[0141] 4. User QoS configuration interface: In the cloud platform user interface, QoS configuration options are provided, allowing users to adjust network priority and resource allocation strategies according to their own needs.

[0142] Through adaptive QoS and traffic engineering, the cloud platform can provide higher quality network services, especially in high-load scenarios. The network performance of key applications and services will be significantly improved, while reducing resource waste and improving overall network efficiency.

[0143] In an exemplary embodiment, the current integration of cloud storage systems and Infiniband networks is often relatively loose, which limits the efficiency of data processing and transmission. Therefore, a deep integration solution is proposed, using the Infiniband network as a bridge to achieve high-speed, low-latency interconnection between cloud storage and computing resources. Specifically:

[0144] 1. Infiniband-aware storage: Develop an Infiniband-aware cloud storage system that enables storage devices to communicate directly with computing nodes over the Infiniband network, avoiding the Ethernet transmission bottleneck in traditional network storage.

[0145] 2. Storage resource virtualization: Cloud storage resources are divided into multiple virtual storage partitions, which match the Infiniband network partitions to achieve storage resource isolation and on-demand allocation in a multi-tenant environment.

[0146] 3. Storage access optimization: Utilizing RDMA (Remote Direct Memory Access) technology, we optimize data storage and access processes, enabling direct memory access between computing nodes and storage resources, significantly reducing data transmission latency and increasing data processing speed.

[0147] 4. Storage and network configuration synchronization: Ensure the synchronization of the cloud storage system configuration with the Infiniband network configuration. For example, when the network partition configuration changes, the corresponding virtual storage partition settings in the storage system are automatically adjusted to maintain system consistency.

[0148] The deep integration of InfiniBand and cloud storage significantly improves data processing and transmission efficiency and reduces data access latency, which is particularly important for scenarios requiring large-scale data processing and rapid data access, such as big data analytics and deep learning. Furthermore, the virtualization and on-demand allocation of storage resources improves cloud platform resource utilization, reduces storage costs, and provides users with a more flexible and efficient data storage solution.

[0149] Obviously, the embodiments described above are only part of the embodiments of the present application, rather than all the embodiments. In order to better understand the above method, the above process is described below in conjunction with the embodiments, but it is not intended to limit the technical solutions of the embodiments of the present application. Specifically:

[0150] This application proposes a multi-tenant Infiniband network management method for a cloud platform, which can automatically configure the Infiniband network, implement a multi-tenant Infiniband network based on partitioning, and provide Infiniband network card configuration for cloud hosts to support tenants' artificial intelligence training, inference and other business scenarios, thereby reducing the overall cost of using the cloud platform.

[0151] The overall idea is as follows:

[0152] 1. Implement the management and configuration of InfiniBand network based on OpenSM, including the management and configuration of IB switches;

[0153] 2. Develop a network plug-in to connect to the OpenStack cloud platform to receive user network creation instructions and send them to the IB network through OpenSM;

[0154] 3. Cloud host network card management can provide a visual way to mount / unmount to the cloud host.

[0155] Specifically:

[0156] (1) Prerequisites: The OFED driver has been installed on each server node, and the OpenSM service is running on the corresponding server node.

[0157] (2) If Figure 3 As shown in the figure, an Infiniband network management module is developed to provide an API interface to the north and to manage and configure Infiniband network devices, network cards, and partitions to the south.

[0158] Specifically, the Infiniband network management module includes an external interface submodule, a configuration management submodule, and a data synchronization submodule, and also supports persisting the configuration to a database.

[0159] 1) External interface submodule: provides API services to the outside world, receives requests, and forwards them to the configuration management submodule;

[0160] 2) Configuration Management Submodule: This module receives requests, converts them into OpenSM configuration files, and sends them to OpenSM. It also persists the configuration information in a database, such as configuring Infiniband partitions and associating network ports with partitions. Synchronization with OpenSM is achieved by connecting to the server via SSH and sending configuration files.

[0161] 3) Data synchronization submodule: To maintain data consistency between the Infiniband network management module and OpenSM, this module is used to synchronize data between the two on a regular and manual basis.

[0162] 4) In terms of functionality, the InfiniBand network management module primarily supports partition configuration and port-partition association. Partition configuration primarily involves Pkey configuration, while network port-partition association primarily involves port list configuration. This module can also be expanded to support other OpenSM configuration operations.

[0163] (3) If Figure 4 As shown, a cloud platform network interface adapter module is developed to receive cloud platform requests and convert them into Infiniband network management module interface format data. The Infiniband network management module interface is called to implement configuration management of the Infiniband network.

[0164] Taking the OpenStack cloud platform as an example, the cloud platform network adapter module must comply with the OpenStack Neutron ML2 specification to implement OpenStack network, subnet, and port business logic. For example, regarding network isolation, Ethernet uses virtual local area networks (VLANs) to isolate subnets, with different subnets having different VLAN IDs. InfiniBand uses partitions, with partitions isolated using private keys. Therefore, each partition must correspond to an OpenStack network concept.

[0165] (4) Cloud host Infiniband network configuration module: Automatically configure the Infiniband network card for the cloud host. In this application, the cloud host Infiniband network card is implemented based on SR-IOV technology. The prerequisite for using this method is to configure the Infiniband physical network card into multiple SR-IOV network cards in advance and configure the Port GUID and Node GUID for each VF.

[0166] The specific process is as follows:

[0167] 1) During the configuration process, you must first schedule an available SR-IOV network card for the cloud host and mount the SR-IOV network card to the cloud host through PCIE passthrough;

[0168] 2) Call the Infiniband network management module to add the VF network card information to the associated partition to implement cloud host support for multi-tenant Infiniband network.

[0169] Optionally, a diagram of allocating partitioned networks to cloud hosts is shown below: Figure 5 shown.

[0170] It should be noted that this application proposes a cloud platform multi-tenant Infiniband network management method that can automatically configure Infiniband networks, implement multi-tenant Infiniband networks based on partitioning, and provide Infiniband network card configuration for cloud hosts to support tenants' business scenarios such as artificial intelligence training and inference, thereby reducing the overall cost of using the cloud platform. Specifically:

[0171] (1) Supports configuring OpenSM through API interfaces instead of configuration files, thereby achieving automated configuration of Infiniband networks;

[0172] (2) Supports persisting Infiniband network configuration to the database, which improves retrieval and configuration efficiency compared to OpenSM's native file-based storage method;

[0173] (3) Supports scheduled and manual synchronization mechanisms to achieve data consistency synchronization between the OpenSM configuration file and the method database. When an OpenSM or underlying device failure occurs, it can be automatically synchronized to the method database and then synchronized to the management platform, improving the overall availability of the method;

[0174] (4) Provide cloud platform docking plug-ins. Without changing the original network interface of the cloud platform, the traditional Ethernet network concept is mapped to the Infiniband network, reducing the workload of the method. At the same time, the scalability of the method is improved. Based on this method, it can be extended to docking heterogeneous cloud platforms such as OpenStack, AWS, and Alibaba Cloud. At the same time, it can be extended to docking Kubernetes based on the plug-in mechanism and extended to scenarios where containers use Infiniband networks.

[0175] (5) By using SR-IOV to split the physical Infiniband into multiple VFs and automatically mounting the VFs to the cloud host, the virtualization of a single physical network card device is achieved, which improves resource utilization and reduces usage costs.

[0176] (6) Through automated partition configuration and associated configuration of cloud host network cards, different Infiniband partitions can be mounted on the network cards of multi-tenant cloud hosts, thereby realizing the configuration of multi-tenant Infiniband networks, meeting the needs of multi-tenant scenarios for Infiniband networks, and expanding the service capabilities of the cloud platform.

[0177] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0178] The embodiment of the present application also provides a configuration device for a high performance computing network. Figure 6 is a structural block diagram of a configuration of a high performance computing network according to an embodiment of the present application, such as Figure 6 As shown, the device includes:

[0179] An acquisition module 602 is configured to acquire network configuration information of a high performance computing network;

[0180] The configuration module 604 is configured to send network configuration information to a subnet manager of the high-performance computing network through the network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information.

[0181] Through the above device, the high-performance computing network is configured by sending network configuration information to the subnet manager through the network management module, thereby realizing the automated configuration of the high-performance computing network, improving the configuration efficiency, and solving the problem that the traditional high-performance computing network configuration method relies on manual operation, resulting in low configuration efficiency.

[0182] Optionally, the acquisition module 602 includes an acquisition unit for acquiring partition configuration information, and / or network topology information, and / or port configuration information, and / or routing configuration information, and / or quality of service configuration information, and / or security configuration information of the high-performance computing network; wherein the network configuration information includes at least one of the following: partition configuration information, network topology information, port configuration information, routing configuration information, quality of service configuration information, and security configuration information.

[0183] Optionally, the device also includes: a sending module, which is used to convert the acquired configuration information into configuration information in a target format through a network adapter module before sending the network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network, wherein the target format is a format corresponding to the application program interface of the network management module, and the configuration information carries the network configuration information; calling the application program interface of the network management module to send the configuration information in the target format to the network management module.

[0184] Optionally, the device further includes: a synchronization module for synchronizing the network configuration information to the target database through the network management module during the process of sending the network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network.

[0185] Optionally, the configuration module 604 includes a first configuration unit, which is used to send a target configuration file to the subnet manager through the network management module, wherein the target configuration file carries network configuration information; or to send network configuration information to the subnet manager through the network management module based on a secure remote access protocol.

[0186] Optionally, the synchronization module is further configured to synchronize the network configuration data in the subnet manager to the target database via the network management module at every preset time interval.

[0187] Optionally, the synchronization module is further configured to synchronize the network configuration data in the subnet manager to the target database through the network management module when a data synchronization instruction is obtained.

[0188] Optionally, the device further includes: a determination module, configured to determine a network status of the high-performance computing network according to network configuration data in the target database.

[0189] Optionally, in the case where the high-performance computing network has N partitions, each of the N partitions corresponds to a network service in the cloud platform, and the network management module is located in the cloud platform.

[0190] Optionally, the device also includes: a processing module for determining the partition of the high-performance computing network selected by the target user and obtaining a target partition identifier when the target user applies to create a target cloud host in the target server; scheduling the target cloud host through the network configuration module so that the target cloud host runs in the network environment corresponding to the target partition identifier.

[0191] Optionally, the processing module includes a first processing unit, which is used to schedule a virtual network card for the target cloud host through the network configuration module; and associate the network card information of the virtual network card to the partition corresponding to the target partition identifier through the network configuration module, so that the target cloud host runs in the network environment corresponding to the target partition identifier.

[0192] Optionally, the first processing unit includes: a first processing sub-unit, configured to determine a virtual network card in an unmounted state from a virtual network card resource pool; and mount the virtual network card to a target cloud host.

[0193] Optionally, the first processing unit includes: a second processing sub-unit, used to determine the port global unique identifier corresponding to the virtual network card; schedule the network management module through the network configuration module to associate the port global unique identifier corresponding to the virtual network card to the partition corresponding to the target partition identifier through the network management module.

[0194] Optionally, the second processing sub-unit is used to associate the port globally unique identifier corresponding to the virtual network card to the partition corresponding to the target partition identifier in the following manner: sending a configuration instruction to the subnet manager through the network management module, so that the subnet manager associates the port globally unique identifier corresponding to the virtual network card to the partition corresponding to the target partition identifier.

[0195] Optionally, the processing module is also used to schedule the target cloud host through the network configuration module so that the target cloud host runs in the network environment corresponding to the target partition identifier, and when the target user requests to modify the network configuration of the network environment corresponding to the target partition identifier, determine the modified configuration information; and send the target instruction to the subnet manager through the network management module so that the subnet manager modifies the network configuration of the partition corresponding to the target partition identifier according to the modified configuration information.

[0196] Optionally, the configuration module 604 includes a second configuration unit for verifying whether each server node of the high-performance computing network has a high-performance network driver installed and whether the target server node has a subnet manager; when each server node of the high-performance computing network has a high-performance network driver installed and the target server node has a subnet manager, network configuration information is sent to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network.

[0197] For the description of the features in the embodiment corresponding to the configuration device of the high-performance computing network, please refer to the relevant description of the embodiment corresponding to the configuration method of the high-performance computing network, and will not be repeated here.

[0198] An embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any of the above-mentioned embodiments of the method for configuring a high-performance computing network.

[0199] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps of any of the above-mentioned embodiments of the method for configuring a high-performance computing network when running.

[0200] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0201] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of any of the above-mentioned embodiments of the method for configuring a high-performance computing network.

[0202] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps in any of the above-mentioned embodiments of the configuration method of the high-performance computing network.

[0203] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0204] The above is a detailed introduction to the configuration method and device of a high-performance computing network, electronic device, storage medium, and computer program product provided by this application. Specific examples are used herein to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method of this application and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to this application, and these improvements and modifications also fall within the scope of protection of the claims of this application.

Claims

1. A method for configuring a high-performance computing network, characterized in that: include: Obtain network configuration information for the high-performance computing network; Sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information; The method further includes: when a target user applies to create a target cloud host in a target server, determining the partition of the high-performance computing network selected by the target user to obtain a target partition identifier; scheduling a virtual network card for the target cloud host through the network configuration module; and associating the network card information of the virtual network card with the partition corresponding to the target partition identifier through the network configuration module, so that the target cloud host runs in the network environment corresponding to the target partition identifier; Among them, the method also includes: collecting network usage data of tenants of the high-performance computing network in real time, and building a network demand module for each tenant based on the network usage data; an intelligent algorithm module based on machine learning, dynamically adjusting the network resources of the high-performance computing network according to the network demand model of each tenant and the real-time status of the network resources of the high-performance computing network.

2. The method for configuring a high performance computing network according to claim 1, wherein: Get network configuration information for the high-performance computing network, including: Obtaining partition configuration information, and / or network topology information, and / or port configuration information, and / or routing configuration information, and / or quality of service configuration information, and / or security configuration information of the high-performance computing network; The network configuration information includes at least one of the following: the partition configuration information, the network topology information, the port configuration information, the routing configuration information, the quality of service configuration information, and the security configuration information.

3. The method for configuring a high performance computing network according to claim 1, wherein: Before sending the network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network, the method further includes: Converting the acquired configuration information into configuration information in a target format through a network adapter module, wherein the target format is a format corresponding to the application program interface of the network management module, and the configuration information carries the network configuration information; The application program interface of the network management module is called to send the configuration information in the target format to the network management module.

4. The method for configuring a high performance computing network according to claim 1, wherein: In the process of sending the network configuration information to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network, the method further includes: The network configuration information is synchronized to the target database through the network management module.

5. The method for configuring a high performance computing network according to claim 1, wherein: Sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network includes: Sending a target configuration file to the subnet manager through the network management module, wherein the target configuration file carries the network configuration information; or The network configuration information is sent to the subnet manager through the network management module based on a secure remote access protocol.

6. The method for configuring a high performance computing network according to claim 1, wherein: The method further comprises: At every preset time interval, the network configuration data in the subnet manager is synchronized to the target database through the network management module.

7. The method for configuring a high performance computing network according to claim 1, wherein: The method further comprises: When a data synchronization instruction is obtained, the network configuration data in the subnet manager is synchronized to the target database through the network management module.

8. The method for configuring a high performance computing network according to claim 6 or 7, wherein: The method further comprises: The network status of the high performance computing network is determined according to the network configuration data in the target database.

9. The method for configuring a high performance computing network according to claim 1, wherein: In the case where the high-performance computing network has N partitions, each of the N partitions corresponds to a network service in the cloud platform, and the network management module is located in the cloud platform.

10. The method for configuring a high performance computing network according to claim 1, wherein: Scheduling a virtual network card for the target cloud host through the network configuration module includes: Determine the virtual network card that is not mounted from the virtual network card resource pool; Mount the virtual network card to the target cloud host.

11. The method for configuring a high performance computing network according to claim 1, wherein: Associating the network card information of the virtual network card with the partition corresponding to the target partition identifier through the network configuration module includes: Determine a globally unique port identifier corresponding to the virtual network card; The network configuration module schedules the network management module so as to associate the port global unique identifier corresponding to the virtual network card with the partition corresponding to the target partition identifier through the network management module.

12. The method for configuring a high performance computing network according to claim 11, wherein: Associating the port globally unique identifier corresponding to the virtual network card with the partition corresponding to the target partition identifier through the network management module includes: A configuration instruction is sent to the subnet manager through the network management module, so that the subnet manager associates the port global unique identifier corresponding to the virtual network card with the partition corresponding to the target partition identifier.

13. The method for configuring a high performance computing network according to claim 1, wherein: Scheduling the target cloud host through a network configuration module so that the target cloud host runs in a network environment corresponding to the target partition identifier, the method further comprising: When the target user requests to modify the network configuration of the network environment corresponding to the target partition identifier, determining modification configuration information; A target instruction is sent to the subnet manager through the network management module, so that the subnet manager modifies the network configuration of the partition corresponding to the target partition identifier according to the configuration modification information.

14. The method for configuring a high performance computing network according to claim 1, wherein: Sending the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network includes: Verify whether each server node of the high-performance computing network has a high-performance network driver installed and whether the target server node has a subnet manager; When each server node of the high-performance computing network is installed with a high-performance network driver and the target server node has a subnet manager, the network configuration information is sent to the subnet manager of the high-performance computing network through the network management module of the high-performance computing network.

15. A configuration device for a high performance computing network, characterized in that: include: An acquisition module, used to obtain network configuration information of a high-performance computing network; A configuration module, configured to send the network configuration information to a subnet manager of the high-performance computing network through a network management module of the high-performance computing network, so that the subnet manager configures the high-performance computing network according to the network configuration information; The device further includes: a processing module for determining the partition of the high-performance computing network selected by the target user and obtaining a target partition identifier when the target user applies to create a target cloud host in the target server; scheduling a virtual network card for the target cloud host through the network configuration module; and associating the network card information of the virtual network card with the partition corresponding to the target partition identifier through the network configuration module, so that the target cloud host runs in the network environment corresponding to the target partition identifier; Among them, the device is also used to collect network usage data of tenants of the high-performance computing network in real time, and build a network demand module for each tenant based on the network usage data; an intelligent algorithm module based on machine learning dynamically adjusts the network resources of the high-performance computing network according to the network demand model of each tenant and the real-time status of the network resources of the high-performance computing network.

16. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for configuring a high-performance computing network as claimed in any one of claims 1 to 14 when executing the computer program.

17. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method for configuring a high-performance computing network according to any one of claims 1 to 14 are implemented.

18. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for configuring a high-performance computing network according to any one of claims 1 to 14 are implemented.

Images