Method of realising intensional intelligence setting in heterogeneous network on the basis of network programming language

Abstract

To provide a method of realizing intentional intelligent setting in a heterogeneous network on the basis of a network programming language.SOLUTION: A method includes the steps of: generating a network policy adapted to Frenetic network programming language specifications by combining a network resource and a state by a service orchestrator on the basis of an intentional keyword; selecting a corresponding domain controller in accordance with a service requirement by a global network controller and distributing the network policy; and compiling the network policy as an OpenFlow flow table by a domain controller that supports the OpenFlow protocol and converting the network policy into a YANG data model by a policy translator via a domain controller that supports NETCONF protocol.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention belongs to the technical field of communication networks, and specifically relates to a method for realizing intelligent setting of user intentions in heterogeneous network environments based on network programming languages.

Background Art

[0002] With the rapid development of information technology, the network has become one of the important infrastructures in modern society. Currently, the network is not only a tool for connecting people and people, and people and information, but also an important driving force for promoting industrial innovation and improving work efficiency. With the rapid expansion and diversification of the network scale, the complexity and heterogeneity of the network structure are increasing day by day, and the requirements of users for network services are also diversified and dynamic.

[0003] Thanks to the proposal of new network technologies such as Software Defined Networking (SDN) and ForCES (Forwarding and Control Element Separation), the conventional network architecture is more flexible and stable than the traditional network architecture. However, the maintenance and management of network lower-level devices still highly depend on the manual settings of network developers. With the expansion of the network scale, the manual setting of the network becomes difficult and inefficient. The stability and accuracy of the entire network are restricted by the risks caused by inevitable human errors, and it also leads to slow service provision and response speed.

[0004] The OpenFlow protocol and the NETCONF protocol are the two most common communication protocols between controllers and devices in the current network architecture. The proposals for both the OpenFlow protocol and the NETCONF protocol provide powerful tools for realizing automated and centralized network management, but the focuses and strengths of the two protocols are different. For example, the OpenFlow protocol enables the controller to directly control the flow tables of network devices, realizing flexible control and management of network traffic, and is more suitable for use in specific scenarios such as data centers or research environments. On the other hand, the NETCONF protocol allows network administrators to securely connect to various functions for remotely configuring network devices and supports functions such as configuration verification and transaction management, with better performance in traditional enterprise networks. The heterogeneous network constructed by combining the OpenFlow and NETCONF protocols can fully utilize the strengths of each protocol, provide flexible network management and control capabilities, and meet diversified service requirements. However, since the deployment and maintenance of heterogeneous networks are relatively complex, there are requirements for the integrated management and intelligent orchestration of services in heterogeneous network environments.

[0005] With the rapid progress of digitalization model changes, users' requirements for network services have become diversified and dynamic. In the conventional network configuration and management methods, it is difficult to obtain and grasp users' service requirements, and it is impossible to customize network services that are also diversified for the service requirements of each network user individual. Intent-based Networking (IBN) is a new emerging technical concept as a brand-new network management and control pattern, aiming to apply more advanced intelligence and expected states instead of the manual process of configuring the network. Because methods of artificial intelligence and machine learning are used, the network software can intelligently consider how to achieve the goal as long as the network administrator defines the execution result or service goal. In the conventional network architecture, by introducing the concept of "intent", if such high-level service goals can be converted into specific configurations of lower-level network element devices, the flexibility and adaptability of the conventional network architecture can be greatly improved, and the requirements of users for diversified services can be better met.

[0006] A network programming language is a domain-specific language (DSL) dedicated to network programming and is used to achieve the configuration and management of network devices. Frenetic is one of the most mature representatives in this technical field and enables the description of network behavior and policies in a way close to natural language. By leveraging the characteristics of Frenetic, which can flexibly describe service capabilities and describe network policies using Python code, and combining with the code generation ability of large language models (LLMs), the programming and configuration processes of network devices can be significantly simplified and accelerated. However, until now, the Frenetic network programming language has only provided support for OpenFlow network devices, which also means that it is difficult for the Frenetic network programming language to function in heterogeneous network environments as an efficient and flexible network management tool.

[0007] As a result, compared with traditional network architectures, the conventional network architecture has seen a huge leap in scale and performance. However, there are still problems such as difficulty in deploying and maintaining heterogeneous networks, low degree of intelligent orchestration, and inability to timely meet users' service requirements.

Summary of the Invention

[0008] In view of the problems existing in the conventional network architecture, the present invention provides a method for setting intentions in heterogeneous networks based on network programming languages. By combining natural language processing, network programming languages, and large language model technologies, flexible design of services targeted at each user intention is realized, while ensuring service provision and response speed, and at the same time improving the diversity of network services and the flexibility of the network. Using automaton theory, a policy translator is designed for the variations of the OpenFlow and NETCONF protocols. In the NETCONF domain, network policies are extracted and converted to generate configuration files for corresponding network element devices, thereby realizing intelligent programming, integrated management, and maintenance of user services in heterogeneous network environments. Also, based on the concept of closed-loop control, the deployed services are monitored and analyzed to design a dynamic service adjustment process, thereby ensuring the QoS of network services.

[0009] A method for realizing intelligent setting of intentions in heterogeneous networks based on network programming languages, which combines intention realization, network programming languages, and policy conversion technologies, comprising: a step in which a global SDN controller acquires heterogeneous network resources and status information; a step in which a user describes network service requirements in the form of voice or text; a step in which an intention analysis engine analyzes the service requirements and extracts intention keywords; a step in which a service orchestrator combines network resources and status based on the intention keywords to generate a network policy that conforms to the Frenetic network programming language specification; a step in which a global network controller selects a corresponding domain controller according to the service requirements and distributes the network policy; A domain controller that supports the OpenFlow protocol compiles the network policy into an OpenFlow flow table, and a domain controller that supports the NETCONF protocol converts the network policy into a YANG data model by a policy translator, An OpenFlow controller distributes the flow table to a lower-layer network element device that supports the OpenFlow protocol, and a NETCONF controller distributes the YANG data model to the corresponding NETCONF network element device, thereby realizing the deployment of the service and starting the closed-loop control.

[0010] The method includes 1) Each domain controller distributes a query command or request to a lower-layer network element device, analyzes the response data returned by the lower-layer network element device, extracts necessary information including device topology information, port status information, fault and alarm information, network function information, and device configuration, and after completing the synchronization of the resource state, the OpenFlow domain controller and the NETCONF domain controller integrally collect the data to a heterogeneous network global SDN controller; 2) A step in which a user inputs an intention in the form of text or voice, where the intention includes the user's network service requirements and a high-level description of the network, representing the network service goals that the user wants to achieve across the heterogeneous network; 3) An intention analysis engine extracts important service information based on the intention input by the user in natural language, converts the intention data into a format processable by a model, extracts intention keywords from the intention data in a format conforming to the specification by a tokenizer, and obtains the service scope desired by the user, the network functions included in the service, and the life cycle information of the service. 4) Input the obtained intended keyword, the current network resource, and the status into the service orchestrator. The service orchestrator generates a network policy that conforms to the Frenetic network programming language specification by means of large language model code generation technology, selects an appropriate domain for the scope of the user service, and distributes the network policy. 5) For domain controllers with different protocols, further process the network policy in different ways. In the OpenFlow domain composed of network element devices that support the OpenFlow protocol, the Frenetic controller compiles the network policy into an OpenFlow flow table. However, in the NETCONF domain composed of network element devices that support the NETCONF protocol, extract and transform the network policy with a policy translator, and further combine it with the lower-level network element information to generate a YANG model file corresponding to the device. 6) The global SDN controller performs a final placement for the instantiated user network service according to the service life cycle described in the user intention, starts the service in production and at the same time starts the closed-loop control module. The global SDN controller monitors the service QoS based on the network status data returned by the domain controller. Once the service can no longer meet the user intention, the global SDN controller starts a request towards the service orchestrator, and the service orchestrator redoes the generation and placement of the service based on the current network resources and status.

[0011] In step 5), in the NETCONF domain, the step of converting the network policy into a YANG model file corresponding to a specific device with a policy translator is 5-1) The extractor extracts a function that describes a network function in a network policy and its processing target based on the data of a deterministic finite state automaton; 5-2) comparing the network function function with a database having a lower layer network element function, and mapping the function in the network policy and its processing target to the function of a specific network element device and the information required for the setting service; 5-3) A policy generator constructed based on a context-free grammar generates a "content generation formula" and a "structure generation formula" based on the data after data conversion and specific network element configuration information, where the content generation formula is used to include data in accurate XML tags, and the structure generation formula is responsible for organizing the overall XML structure and ensuring the accurate relationship between each part; 5-4) outputting lower layer network element configuration information, that is, a YANG model file.

[0012] The beneficial effects of the present invention are as follows. The present invention discloses a method for intelligent configuration of intentions in heterogeneous networks based on network programming languages, realizes intelligent conversion from user service requirements to device configuration, simplifies the service realization process, and improves network performance. Specifically, by combining intention realization, network programming languages, and large language model technologies, and designing an intelligent retranslation method from user intentions to network policies, flexible service design for each user is realized. In addition, based on automaton theory and context-free grammar, a policy translator that converts network policies conforming to the Frenetic network programming language specification into network element device configurations is designed, solving the problem that the Frenetic network programming language and the NETCONF protocol are not compatible, and realizing intelligent orchestration, integrated management, and maintenance of user services in heterogeneous network environments. Finally, based on the concept of closed-loop control, by monitoring and analyzing the deployed services and designing a dynamic service adjustment process, the QoS of network services is guaranteed. The embodiments of the present invention can be applied to the network architecture of network service providers, solving the problems that the heterogeneous networks still existing in the existing network architecture are difficult to deploy and maintain, have a low degree of intelligent orchestration, and cannot meet user service requirements in a timely manner, greatly improving the diversity of network services, as well as the flexibility, accuracy, and stability of the network, and effectively ensuring service provision and response speed.

Brief Description of the Drawings

[0013]

Figure 1

[0014]

Figure 2

Embodiments for Carrying Out the Invention

[0015] In the following, the present invention will be further described with reference to the drawings and embodiments. Embodiment

[0016] A certain network service provider manages a heterogeneous network architecture composed of multiple domains, and each domain is composed of devices supporting the OpenFlow protocol or the NETCONF protocol. The network service provider plans to implement intent acquisition, intelligent orchestration, and integrated placement to perform closed-loop control when a network user issues service requirements, and the method is as shown in FIG. 1.

[0017] 1-1) Before providing a service to a network user, the network service provider needs to clarify the current network resources and status. The domain controller establishes a connection with lower-layer network element devices and network links through the OpenFlow protocol or the NETCONF protocol, and distributes query commands or requests. After receiving the request, the lower-layer network element device returns device topology information, port status information, fault and alarm information, network function information, and device configuration data. After completing the synchronization of the resource status, the domain controller aggregates the data to the global SDN controller.

[0018] 1-2) The network service provider provides an interface to the user, and the user expresses his / her intent in the form of text or voice. Assume that what the user inputs in the form of natural language is "Please prohibit all hosts in the network from accessing the entertainment website on business days." 1-3) After receiving the intent from the user, the intent analysis engine first converts the input natural language into a format that conforms to the specification. For example, { “role”:“user”:, "content": "Please prohibit access to the entertainment websites of all hosts in the network on business days.": } Next, the tokenizer extracts the intended keywords from the intended data in a format that conforms to the specifications, including the service scope desired by the user, the network functions included in the service, and the life cycle of the service. For example, Service scope desired by the user: All hosts; Network functions included in the service: Prohibit access to entertainment websites on business days; Service life cycle: Business days.

[0019] 1 - 4) In the service orchestrator, using the intended keywords obtained through the above processing, as well as the lower - level network resources and states collected by the global SDN controller as input, a network policy that conforms to the Frenetic network programming language specifications is generated by the fine - tuned large - language model. For example, import frenetic from frenetic.syntax import * class MyApp(frenetic.App): p1 = IP4DstEq(addr1,mask) / / Matches packets with the destination IP address addr1 p2 = IP4DstEq(addr2,mask) …… pn = IP4DstEq(addrn,mask) / / addr is the IP address of the entertainment website that needs to be blocked selectPacket = Or([p1, p2,..., pn]) Drop(selectPacket) app = MyApp() app.start_event_loop().

[0020] 1 - 5) Determine the domain for placing user services based on user information and intentions, and further distribute the generated network policy to the corresponding domain controller.

[0021] In an OpenFlow domain composed of OpenFlow network element devices, the network policy is directly compiled as an OpenFlow flow entry by the Frenetic controller. It is as follows, where 220.181.38.148 is one of the specific IP addresses of addr in the above network policy. Flow Table Entry: Match Fields: - IPsrc != 220.181.38.148 - IPdst = 220.181.38.148 Instructions: - Drop.

[0022] In a NETCONF domain composed of NETCONF network element devices, the network policy is further converted into lower - level network element configuration information by a policy translator. As shown in Figure 2, the specific process is as follows. 1 - 5 - 1) The extractor analyzes the network policy based on the data of a deterministic finite - state automaton, extracts the function describing the network function in the network policy and its operation target. Specifically, it includes IP4DstEq, the drop function, device IP, and port information.

[0023] 1 - 5 - 2) Compare the network function function with a database with lower - level network element functions, and convert the function and its operation target in the network policy into information required for the functions and setting services of specific network element devices. For example, <access-lists> <access-list> <name>Block Entertainment< / name> <aces> <ace> <name>Deny Entertainment Site 1< / name> <matches> <ipv4> <destination-ip-address> 220.181.38.148< / destination-ip-address> < / ipv4> < / matches> <actions> <forwarding>drop< / forwarding> < / actions> < / ace> <!-- Add more ACEs for other entertainment sites --> < / aces> < / access-list> < / access-lists> .

[0024] 1-5-3) The policy generator constructed based on the context-free grammar generates "content generation expressions" and "structure generation expressions" based on the data after being converted by the data converter and the configuration information of specific network elements. The content generation expressions are used to include the data in accurate XML tags, and the structure generation expressions are used to packetize other tags. Finally, based on the content of both, the lower-level network element configuration information, that is, the YANG model file, is output.

[0025] 1-6) Based on the service life cycle described in the user intention, the instantiated network service is deployed. For the OpenFlow domain, the domain controller updates the OpenFlow flow table. For the NETCONF domain, the domain controller distributes the YANG file model by NETCONF. When the network service goes into production, at the same time, the global SDN controller activates the closed-loop control module, monitors and analyzes the network resources and status based on the port status information, faults and alarm information returned by the domain controller, and determines whether the user service QoS is guaranteed. Once the service can no longer meet the user intention, the global SDN controller issues a request to the service orchestrator, and the service orchestrator re-performs the generation and deployment of the service based on the current network resources and status.

[0026] The embodiments described above can be further combined or switched, and the embodiments only illustrate the preferred examples of the present invention, and do not limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various changes and improvements made by those skilled in the art to the technical solutions of the present invention all belong to the protection scope of the present invention. The protection scope of the present invention is shown by the appended patent claims and all equivalent technical solutions thereof.

Claims

1. Combining the realization of intent, network programming language, and policy conversion technology, Steps for the global SDN controller to obtain heterogeneous network resources and status information, Steps for the user to describe network service requirements in the form of voice or text, Steps for the intent analysis engine to analyze service requirements and extract intent keywords, Steps for the service orchestrator to combine network resources and status based on intent keywords to generate a network policy that conforms to the Frenetic network programming language specification, Steps for the global network controller to select the corresponding domain controller according to service requirements and distribute the network policy, Steps for the domain controller supporting the OpenFlow protocol to compile the network policy into an OpenFlow flow table, and for the domain controller supporting the NETCONF protocol to convert the network policy into a YANG data model by a policy translator, Steps for the OpenFlow controller to distribute the flow table to the subordinate network element devices supporting the OpenFlow protocol, and for the NETCONF controller to distribute the YANG data model to the corresponding NETCONF network element devices, thereby realizing service deployment and activating closed-loop control. A method for realizing intelligent configuration of intent in heterogeneous networks based on a network programming language, characterized by including the above steps.

2. 1) Each domain controller distributes query commands or requests to subordinate network element devices, analyzes the response data returned by the subordinate network element devices, extracts necessary information including device topology information, port status information, fault and alarm information, network function information, and device configuration, and after completing resource status synchronization, the OpenFlow domain controller and the NETCONF domain controller integratively summarize the data to the heterogeneous network global SDN controller. 2) A step in which the user inputs an intention in the form of text or voice, where the intention includes the user's network service requirements and a high-level description of the network, and represents the network service goals the user wishes to achieve across heterogeneous networks. 3) A step in which the intention analysis engine extracts important service information based on the intention input by the user in natural language form, converts the intention data into a format processable by a model, extracts intention keywords from the intention data in a format conforming to the specification by a tokenizer, and obtains the service scope desired by the user, the network functions included in the service, and the life cycle information of the service. 4) Input the obtained intention keywords, the current network resources and status into the service orchestrator, and the service orchestrator generates a network policy conforming to the Frenetic network programming language specification by large language model code generation technology, selects an appropriate domain for the user service scope and distributes the network policy. 5) For different domain controllers of protocols, further process the network policy in different ways. In the OpenFlow domain composed of network element devices supporting the OpenFlow protocol, the Frenetic controller compiles the network policy as an OpenFlow flow table, but in the NETCONF domain composed of network element devices supporting the NETCONF protocol, the network policy is extracted and converted by a policy translator, and further combined with lower-level network element information to generate a YANG model file corresponding to the device. 6) The global SDN controller performs a final placement for the instantiated user network service according to the service life cycle described in the user intention, puts the service into production, and at the same time starts the closed-loop control module. The global SDN controller monitors the service QoS based on the network state data returned by the domain controller. Once the service can no longer meet the user intention, the global SDN controller starts a request towards the service orchestrator, and the service orchestrator re-performs the generation and placement of the service based on the current network resources and status. The method according to claim 1, characterized by including the above steps.

3. In step 5), in the NETCONF domain, the step of converting the network policy into a YANG model file corresponding to a specific device by the policy translator is as follows: 5-1) An extractor extracts a function describing the network function in the network policy and its processing target based on the data of the deterministic finite state automaton. 5-2) Compare the network function function with a database having lower-layer network element functions, and map the function and its processing target in the network policy to the functions of specific network element devices and the information required for the configuration service. 5-3) A policy generator constructed based on context-free grammar generates a "content generation formula" and a "structure generation formula" based on the data after data conversion and the specific network element configuration information. The content generation formula is used to include data in accurate XML tags, and the structure generation formula is responsible for organizing the overall XML structure and ensuring the correct relationship between each part. 5-4) Output lower-layer network element configuration information, that is, a YANG model file. The method according to claim 1, characterized by including the above steps.

Images