A network slicing test method and terminal for white-box switches

Abstract

This invention discloses a network slicing testing method and terminal for white-box switches. First, a white-box switch model and a link model are constructed. Based on the constructed models and the actual connection conditions, a twin network model is built. Simultaneously, the twin network model is monitored. Based on the resource predictions of the white-box switch and physical links, a corresponding network slicing scheme is performed. The network slicing scheme is then deployed, optimized, and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output. This method enables network slicing testing, and the combined deployment and optimization process significantly shortens the development cycle of slicing schemes.

Description

Technical Field

[0001] This invention relates to the technical field of network testing, and in particular to a network slicing testing method and terminal for a white-box switch. Background Technology

[0002] With the in-depth construction of smart grids and distribution network systems, new power systems urgently need high-bandwidth, high-reliability, and low-latency mobile communication networks. Network slicing, as a key technology of 5G (5th Generation Mobile Communication Technology), leverages Network Function Virtualization (NFV) and Software Defined Networking (SDN) technologies to enable networks to meet the differentiated needs of various distribution network services.

[0003] Currently, with the rapid development of the power Internet of Things (IoT), 5G distribution network slicing services are increasing dramatically, which also places higher demands on the resource optimization capabilities of network slicing. To meet the service needs of different services, resources in the power network need to be rationally allocated. As a key device for future network openness and innovation, white-box switches adopt a network architecture that supports open interfaces and network programmability, aligning with the needs of distribution network services and future network development. However, current power systems composed of white-box switches still lack a unified intelligent SDN management platform, making it impossible to directly deploy artificial intelligence algorithms based on reinforcement learning and other technologies within the power system. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a network slicing test method and terminal for white-box switches, which can perform network slicing tests and shorten the development cycle of slicing solutions.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A network slicing testing method for white-box switches includes the following steps:

[0007] Collect data from white-box switches and physical links, store the collected data in a database, and construct white-box switch models and physical link models based on the database;

[0008] Construct a twin network model of the physical network based on the white-box switch model and the physical link model;

[0009] The twin network model is monitored, and the network slicing scheme is predicted according to the resources of the white-box switches and physical links. The network slicing scheme is deployed, optimized and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output.

[0010] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows:

[0011] A network slicing test terminal for a white-box switch includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the network slicing test method for a white-box switch described above.

[0012] The beneficial effects of this invention are as follows: First, a white-box switch model and a link model are constructed. Based on the constructed models and the actual connection situation, a twin network model is built. Simultaneously, the twin network model is monitored. Based on the resource predictions of the white-box switches and physical links, a corresponding network slicing scheme is proposed. The network slicing scheme is then deployed, optimized, and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output. In this way, network slicing testing can be performed, and the deployment and optimization process significantly shortens the development cycle of the slicing scheme. Attached Figure Description

[0013] Figure 1 This is a flowchart of a network slicing test method for a white-box switch according to an embodiment of the present invention;

[0014] Figure 2 This is a schematic diagram of a network slicing test terminal for a white-box switch according to an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of the data representation of a white-box switch according to an embodiment of the present invention;

[0016] Figure 4 This is a schematic diagram of the data representation of the physical link in an embodiment of the present invention;

[0017] Figure 5 This is a schematic diagram illustrating the deployment, optimization, and testing schemes of the network slicing solution according to an embodiment of the present invention;

[0018] Figure 6 This is a schematic diagram illustrating the construction of a white-box switch and link model according to an embodiment of the present invention.

[0019] Label Explanation:

[0020] 1. A network slicing test terminal for a white-box switch; 2. Memory; 3. Processor. Detailed Implementation

[0021] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0022] Please refer to Figure 1 This invention provides a network slicing testing method for white-box switches, including the following steps:

[0023] Collect data from white-box switches and physical links, store the collected data in a database, and construct white-box switch models and physical link models based on the database;

[0024] Construct a twin network model of the physical network based on the white-box switch model and the physical link model;

[0025] The twin network model is monitored, and the network slicing scheme is predicted according to the resources of the white-box switches and physical links. The network slicing scheme is deployed, optimized and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output.

[0026] As described above, the beneficial effects of this invention are as follows: First, a white-box switch model and a link model are constructed. Based on the constructed models and the actual connection situation, a twin network model is built. Simultaneously, the twin network model is monitored. Based on the resource predictions of the white-box switches and physical links, a corresponding network slicing scheme is developed. The network slicing scheme is then deployed, optimized, and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output. In this way, network slicing testing can be performed, and the deployment and optimization process significantly shortens the development cycle of the slicing scheme.

[0027] Furthermore, the data collected from the white-box switch and physical link includes:

[0028] Collect basic attribute information, port information, flow table entry information, performance information, and log information of the white-box switch;

[0029] Collect basic attribute information, link performance information, and log information of the physical link.

[0030] As described above, collecting data from white-box switches and physical links facilitates the establishment of subsequent models.

[0031] Furthermore, storing the collected data in the database includes:

[0032] The basic attribute information of the white-box switch and the physical link, the port information of the white-box switch, and the flow table entry information are stored in a structured database.

[0033] The log information and performance information of the white-box switch and the physical link are stored in a distributed file database.

[0034] As described above, structured databases refer to data that share a common characteristic. Distributed file databases are mainly used to store network log information and error logs, thus classifying and storing data into corresponding databases.

[0035] Furthermore, constructing a white-box switch model and a physical link model based on the database includes:

[0036] Three-dimensional models of white-box switches and physical links are established respectively. Based on the database, the collected white-box switch data is entered into the three-dimensional model of the white-box switch, and the collected physical link data is entered into the three-dimensional model of the physical link.

[0037] As described above, constructing a 3D model of the white-box switch and physical link, and setting its attributes, facilitates browsing of the white-box switch and physical link.

[0038] Furthermore, based on the database, a white-box switch model and a physical link model are constructed, which then includes:

[0039] The collected data on white-box switches and physical links are iterated and optimized to improve the white-box switch model and physical link model.

[0040] As described above, iterative optimization based on the actual data collected from the twin database makes the constructed digital twin model more accurate.

[0041] Furthermore, the network slicing schemes corresponding to resource predictions of white-box switches and physical links include:

[0042] Based on the collected data from white-box switches and physical links, resource requirements are predicted, and a network slicing scheme for virtual network functions is set up according to the predicted resource requirements.

[0043] As described above, based on historical data collected from white-box switches and physical links, future resource requirements can be predicted, allowing for the advance design of VNF deployment, scheduling, and migration schemes to ensure normal service operation. Network slicing schemes based on resource prediction can guarantee the real-time nature and effectiveness of policies.

[0044] Furthermore, the process of collecting data from the white-box switch and physical link includes receiving network service requests;

[0045] Deploying and performing performance tests on the network slicing scheme includes:

[0046] In the twin network model, a deployment scheme for virtual network functions is generated based on the network slicing scheme;

[0047] Determine whether the deployment scheme can meet all service requirements in the network service request. If so, perform a performance evaluation of the white-box switch and physical link. Otherwise, generate a scheduling and migration scheme for the virtual network function until all service requirements in the network service request are met, and then perform a performance evaluation of the white-box switch and physical link.

[0048] As described above, based on the constructed twin network model, a specific VNF deployment scheme can be designed using reinforcement learning. This scheme specifies which white-box switch the VNF should be deployed on, while consuming corresponding physical resources. When the deployment scheme cannot meet the needs of all services, a reasonable VNF scheduling and migration scheme can be designed to meet the needs of different services, greatly shortening the development cycle of the slicing scheme.

[0049] Furthermore, the deployment scheme for generating virtual network functions in the twin network model based on the network slicing scheme includes:

[0050] Based on the reliability requirements of network service requests, a deployment scheme for virtual network functions that meets the reliability requirements is generated in the twin network model through reinforcement learning according to the network slicing scheme.

[0051] As described above, for some services with high reliability requirements, high-reliability deployment of VNFs can be achieved through reinforcement learning, deep reinforcement learning, and other solutions, thereby improving the quality of user service experience.

[0052] Furthermore, the performance evaluation of the white-box switch and physical link includes:

[0053] Network performance is calculated based on the performance of the white-box switch and the physical link. The network performance includes at least one of the following: end-to-end latency, bandwidth, jitter, packet loss rate, and SLA metrics of the slice service function chain.

[0054] As can be seen from the above description, performance evaluation of white-box switches and physical links through network performance calculations can ensure the rationality of the performance evaluation.

[0055] Please refer to Figure 2 Another embodiment of the present invention provides a network slicing test terminal for a white-box switch, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the network slicing test method for a white-box switch described above.

[0056] The network slicing testing method and terminal for white-box switches described above are applicable to slicing testing of software-defined networks using white-box switches, enabling control and management of the white-box switches and shortening the development cycle of slicing solutions. The following detailed implementation methods illustrate these methods:

[0057] Example 1

[0058] Please refer to Figure 1 A network slicing testing method for white-box switches, comprising the following steps:

[0059] S1. Collect data from white-box switches and physical links, and store the collected data in a database. Construct a white-box switch model and a physical link model based on the database.

[0060] S11. Receive network service request. In this embodiment, a service request sent by the user is received, and authorization of the user is completed simultaneously.

[0061] S12. Collect basic attribute information, port information, flow table entry information, performance information and log information of the white-box switch, and collect basic attribute information, link performance information and log information of the physical link.

[0062] Specifically, the Ryu controller collects basic attribute data, flow table entry data, port data, performance data, log data, etc. of white-box switches and physical links.

[0063] Please refer to Figure 3 The basic attribute information of a white-box switch includes the switch identifier, IP address, subnet mask, gateway, switch operating status, and connection status; port information includes the switch port number, the amount of data transmitted through the port, and the amount of data received through the port; flow table entry information includes the source IP address, source MAC address, destination IP address, destination MAC address, packet size, and the current action of the switch on the packet, which includes forwarding or dropping; performance information includes transmission latency, computation latency, and transmission rate; and log information includes the specific usage of the switch over a period of time.

[0064] Please refer to Figure 4 The basic attribute information of the physical link includes the link identifier, link operating status, and link connection status; the link performance information includes the link transmission latency, transmission bandwidth, link jitter, and link packet loss rate; the log information includes the specific usage of the link over a period of time.

[0065] S13. Store the basic attribute information of the white-box switch and the physical link, the port information of the white-box switch and the flow table entry information into a structured database, and store the log information and performance information of the white-box switch and the physical link into a distributed file database.

[0066] Specifically, the databases include structured databases and HDFS (Distributed File System) databases, which together are referred to as the digital twin database of the Software-Defined Networking (SDN) slicing test platform.

[0067] Structured databases refer to databases where data shares a common characteristic and can be logically expressed using a two-dimensional table structure, meaning they possess a defined data structure. Structured databases are primarily used to store basic attribute information, port information, and flow table entries for switches.

[0068] HDFS databases are primarily used to store network log information and error logs. When the network is functioning normally, the HDFS database records network log information, allowing operators and developers to easily view the logs and understand the network's operational mechanisms. When illegal errors or failures occur, the database records the cause of the error, helping people quickly resolve the problem. The database needs to achieve the performance requirement of storing detailed and accurate network operation logs to facilitate platform development and maintenance.

[0069] S14. Establish three-dimensional models of the white-box switch and the physical link respectively. Based on the database, input the collected white-box switch data into the three-dimensional model of the white-box switch and input the collected physical link data into the three-dimensional model of the physical link.

[0070] For details, please refer to Figure 6 The 3D model is constructed using a 3D visualization module. The 3D model of the switch mainly includes the switch backplane, processor, memory, ports, interfaces, and power supply, and is built using 3ds Max software. The white-box switch 3D model is built in 3ds Max, converted to an .fbx file, and imported into the Unity 3D display engine. Appropriate textures, materials, and animations are set for the model, and actions such as movement, rotation, scaling, selection, and connection are imported into Unity 3D for easy viewing of the white-box switch.

[0071] The system sets attributes for the modules through the attribute setting module. Each model's attribute setting module has an interface with the platform's digital twin database, and the system automatically enters the basic attribute data of the white-box switch collected from the database.

[0072] Logs are entered through the log reporting module. For each white-box switch and each network cable or adapter cable, there is also an interface between the log reporting module and the digital twin database. The system will automatically record the logs of the switch and physical links at every moment, which makes it easy to trace the life cycle of the equipment when the equipment fails or ages.

[0073] Performance testing is performed through the performance testing module. Since the Ryu controller can collect information such as the transmission latency, calculation latency, and transmission rate of the white-box switch, as well as the transmission latency, transmission bandwidth, link jitter, and link packet loss rate of the link, the platform can monitor the performance of each switch device and link in real time.

[0074] S15. Iterate through the collected data of white-box switches and physical links, and optimize the white-box switch model and physical link model.

[0075] In this embodiment, optimization is performed through a model optimization module. Due to the certain delay in data collection on the platform, in order to ensure the accuracy of performance, this invention proposes a model optimization module, which uses GRU, GNN, and Transformer to predict performance in advance and iteratively optimizes it based on the actual data collected from the twin database, making the constructed digital twin model more accurate.

[0076] S2. Construct a twin network model of the physical network based on the white-box switch model and the physical link model.

[0077] Specifically, by calling the white-box switch model and link model, a twin network model that is completely consistent with the physical network can be constructed.

[0078] Please refer to Figure 5 The twin network model includes an environmental monitoring module, which comprises two parts: 3D twin environmental monitoring and real-time video monitoring. The 3D twin environmental monitoring refers to the twin network model constructed by the system using white-box switch and link models. This model has an interface with the performance evaluation module, enabling the platform to not only monitor the performance of each white-box switch and link in real time, but also to evaluate the performance of the entire network. Real-time video monitoring refers to the platform installing multiple cameras in the actual switch operating environment to observe the switch's operating status at any time, preventing fires or other safety accidents caused by electrical leakage.

[0079] S3. Monitor the twin network model, predict the network slicing scheme based on the resources of the white-box switch and physical links, deploy, optimize and test the performance of the network slicing scheme, and output the corresponding network slicing scheme when the network slicing scheme reaches the preset performance.

[0080] S31. Based on the collected data of white-box switches and physical links, predict resource requirements, and set up a network slicing scheme for virtual network functions according to the predicted resource requirements.

[0081] For details, please refer to Figure 5 The twin network model also includes a network slice resource prediction module. This module allows the platform to predict future resource requirements based on historical data collected from white-box switches and physical links, thereby enabling the pre-design of Virtual Network Functions (VNF) deployment, scheduling, and migration schemes to ensure normal service operation. Network slicing schemes based on resource prediction can guarantee the real-time performance and effectiveness of policies.

[0082] S32. In the twin network model, generate a deployment scheme for virtual network functions based on the network slicing scheme.

[0083] For details, please refer to Figure 5 The twin network model also includes a network slice VNF deployment module. In the SDN slicing test platform for white-box switches, the network slice VNF deployment scheme refers to the platform's ability to design a specific VNF deployment scheme based on the constructed twin network model through reinforcement learning when slice service demand arrives. This scheme specifies which white-box switch the VNF should be deployed on, while consuming the corresponding physical resources.

[0084] In some embodiments, based on the reliability requirements of network service requests, a deployment scheme for virtual network functions that meet the reliability requirements is generated in the twin network model through reinforcement learning according to the network slicing scheme.

[0085] For details, please refer to Figure 5 The twin network model also includes a network slice reliability assurance module. For some eMBB services with high reliability requirements, the platform can achieve high-reliability deployment of VNFs through reinforcement learning, deep reinforcement learning and other schemes, thereby improving the quality of user service experience.

[0086] S33. Determine whether the deployment scheme can meet all service requirements in the network service request. If so, perform a performance evaluation of the white-box switch and physical link. Otherwise, generate a scheduling and migration scheme for the virtual network function until all service requirements in the network service request are met, and then perform a performance evaluation of the white-box switch and physical link.

[0087] For details, please refer to Figure 5The twin network model includes a network slice VNF scheduling and migration module. This means that when the VNF deployment scheme designed by the platform cannot meet the QoS requirements of all services, a reasonable VNF scheduling and migration scheme can be designed to meet the QoS requirements of different services. After the scheme is verified in the twin network model, it is then distributed to each Ryu controller.

[0088] The network performance is calculated based on the performance of the white-box switch and the physical link. The network performance includes at least one of the following: end-to-end latency, bandwidth, jitter, packet loss rate, and SLA metrics of the slice service function chain.

[0089] For details, please refer to Figure 5 The twin network model also includes a performance evaluation module: the network performance can be calculated based on the performance of white-box switches and links. In this embodiment, the network performance mainly includes end-to-end latency, bandwidth, jitter, packet loss rate and SLA metrics of the Slice Service Function Chain (SFC).

[0090] Please refer to Figure 5 The twin network model also includes a confirmation and distribution module, which means that the system continuously evaluates the effectiveness of the optimization scheme based on the functional module and the performance evaluation module. If it is determined that the optimization model in the functional module has achieved the expected performance, the network will also simulate in the module to determine whether the QoS and SLA of the service have been optimized or meet the user requirements before distributing to each Ryu controller. The Ryu controller will control the transmission of data packets of the white box switch based on the optimization scheme of the functional model.

[0091] This embodiment fully considers the lack of a unified SDN intelligent management and control platform in the power system, enabling intelligent management and control of the distribution network. Furthermore, the slicing scheme effectively improves network resource utilization and service acceptance. The designed modules enable rapid construction of digital twin models and ensure their accuracy, while also significantly shortening the development cycle of the platform's AI intelligent slicing solution.

[0092] Example 2

[0093] Please refer to Figure 2 A network slicing test terminal 1 for a white-box switch includes a memory 2, a processor 3, and a computer program stored in the memory 2 and executable on the processor 3. When the processor 3 executes the computer program, it implements the various steps of the network slicing test method for a white-box switch according to Embodiment 1.

[0094] In summary, the network slicing testing method and terminal for white-box switches provided by this invention first constructs a digital twin network model based on the existing white-box switch model and link model, along with the actual connection conditions. Simultaneously, the performance evaluation module evaluates the overall network performance in real time. If, at a certain moment or time slot, the existing slicing scheme no longer meets the user's QoS or SLA requirements, the system will call the functional model in the functional module to optimize the slicing scheme and submit it to the performance evaluation module for evaluation. After confirmation by the confirmation and distribution module that the user's requirements are met, the scheme is then distributed to each Ryu controller. If the scheme still does not meet the requirements, it returns to the functional module for iterative optimization until a satisfactory scheme is obtained. Therefore, the slicing scheme can effectively improve network resource utilization and service acceptance. The designed modules enable rapid construction of the digital twin model and ensure its accuracy, while also significantly shortening the development cycle of the platform's AI intelligent slicing solution.

[0095] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A network slicing test method for a white-box switch, characterized in that, Including the following steps: Collect data from white-box switches and physical links, store the collected data in a database, and construct white-box switch models and physical link models based on the database; Construct a twin network model of the physical network based on the white-box switch model and the physical link model; The twin network model is monitored, and the network slicing scheme is predicted according to the resources of the white-box switches and physical links. The network slicing scheme is deployed, optimized and its performance is tested. When the network slicing scheme reaches the preset performance, the corresponding network slicing scheme is output.

2. The network slicing test method for a white-box switch according to claim 1, characterized in that, The data collected from the white-box switch and physical link includes: Collect basic attribute information, port information, flow table entry information, performance information, and log information of the white-box switch; Collect basic attribute information, link performance information, and log information of the physical link.

3. The network slicing test method for a white-box switch according to claim 2, characterized in that, The process of storing the collected data in the database includes: The basic attribute information of the white-box switch and the physical link, the port information of the white-box switch, and the flow table entry information are stored in a structured database. The log information and performance information of the white-box switch and the physical link are stored in a distributed file database.

4. The network slicing test method for a white-box switch according to claim 1, characterized in that, Constructing a white-box switch model and a physical link model based on the database includes: Three-dimensional models of white-box switches and physical links are established respectively. Based on the database, the collected white-box switch data is entered into the three-dimensional model of the white-box switch, and the collected physical link data is entered into the three-dimensional model of the physical link.

5. The network slicing test method for a white-box switch according to claim 1, characterized in that, Based on the database, a white-box switch model and a physical link model are constructed, followed by: The collected data on white-box switches and physical links are iterated and optimized to improve the white-box switch model and physical link model.

6. The network slicing test method for a white-box switch according to claim 1, characterized in that, Network slicing schemes based on resource prediction of white-box switches and physical links include: Based on the collected data from white-box switches and physical links, resource requirements are predicted, and a network slicing scheme for virtual network functions is set up according to the predicted resource requirements.

7. The network slicing test method for a white-box switch according to claim 1, characterized in that, Before collecting data from white-box switches and physical links, the process includes: receiving network service requests; Deploying and performing performance tests on the network slicing scheme includes: In the twin network model, a deployment scheme for virtual network functions is generated based on the network slicing scheme; Determine whether the deployment scheme can meet all service requirements in the network service request. If so, perform a performance evaluation of the white-box switch and physical link. Otherwise, generate a scheduling and migration scheme for the virtual network function until all service requirements in the network service request are met, and then perform a performance evaluation of the white-box switch and physical link.

8. A network slicing test method for a white-box switch according to claim 7, characterized in that, In the twin network model, the deployment scheme for generating virtual network functions based on the network slicing scheme includes: Based on the reliability requirements of network service requests, a deployment scheme for virtual network functions that meets the reliability requirements is generated in the twin network model through reinforcement learning according to the network slicing scheme.

9. A network slicing test method for a white-box switch according to claim 7, characterized in that, The performance evaluation of white-box switches and physical links includes: Network performance is calculated based on the performance of the white-box switch and the physical link. The network performance includes at least one of the following: end-to-end latency, bandwidth, jitter, packet loss rate, and SLA metrics of the slice service function chain.

10. A network slicing test terminal for a white-box switch, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements each step of the network slicing test method for a white-box switch as described in any one of claims 1 to 9.

Images