Network Management Intent -One of IBN Use Cases

With the rapid evolution of networks, such as the emergence of the sixth generation (6G), we have entered a new digital era that is ubiquitously connected by highly heterogeneous and dynamic network infrastructure. And various network applications are predicted to appear more in future communication networks. With these complex network scenarios, services, and uncountable degrees of freedom in future communication networks, the complexity of network management will increase drastically. Traditional network management methods are insufficient to keep up with the growing requirements. Firstly, there exists high complexity and difficulty to manage a large amount of infrastructures due to high labor cost, high error probability, low management efficiency. Secondly, traditional network management lacks closed-loop control, which can not find and repair faults in time. To overcome these challenges, some novel network models have been proposed, such as NFV and SDN. However, it still faces many novel challenges: Tight coupling of network and service applications: Traditional network management methods cannot solve the diverse cross-domain service requirements in future communication networks High configuration complexity and scaling cost: Future communication networks will have a huge network scale, different kinds of network resources, and constantly changing topology. This will result in a high network configuration complexity and cannot be configured in a timely and efficient manner. A smart and simple network management architecture is required for rapid deployment of network service requirements Fragile performance provision and policy robustness: In future communication networks, it need to continuously improve the network system model based on the experienced knowledge of administrators. And real-time verification and feedback mechanisms are required for network runtime. Lack of full life cycle verification of policy resilience: Manual management has several shortcomings, such as high error probability, long fault location time, and expensive labor costs. Fortunately, full life cycle verification can reduce the failure recovery time, monitor network operation, and maintain the whole process while enhancing the accuracy of policy configuration and the robustness of policy implementation. At the same time, with the exponential growth of network devices, network administrators need to put in tremendous effort to manage the policies that affect the services of these devices. While in recent years the network management methods have been gradually automated, there are still many procedures that must be accomplished under the strict supervision of administrators, resulting in high error probability. Moreover, current network management is at a level too low to entirely eliminate the requirements to customize each solution for a specific device or protocol in use. The emergence of the IBN, as defined in RFC 9315, has the potential to compensate for the limitations of the current network management methods. The intent is a high-level abstraction of policy, and the IBN is a way to manage the network through intent-driven rather than a low-level configuration. When the user specifies a high-level goal (intent), the network automatically converts that goal into policies and automatically deploys these policies throughout the network.

IBN: Intent based network IBNM: Intent based network management DQN: Deep Q network

Driven by the above requirements and challenges, the intent based network management (IBNM) is proposed. IBNM aims to transition from a fully static manual network management to a fully dynamic autonomous intent based network management. In IBNM, users express their service requirements or goals in a declarative manner without paying attention to how the network achieves them. The architecture is shown as Fig 1, which includes the application layer, the intent-enabled layer and the infrastructure layer. The application layer collects intents from various users and applications, and provides a number of programmable network management services. The intent-enabled layer consists of the intent translation module, intelligent policy mapping module, and intent guarantee module, whose functions are to build a bridge between the application layer and the infrastructure layer. Heterogeneous physical devices are deployed in the infrastructure layer. This layer can execute management instructions from the intent-enabled layer and upload underlying network situation information to the intent-enabled layer. Information interaction between different layers is done through different interfaces, such as the northbound and southbound interfaces.

+----------------------------------------+ | Application Layer | +-------------+---------^----------------+ Intent Ingestion| | Northbound Interface +-------------+---------v----------------+ | | Intent-enabled Layer| | +-----------+-------+ +-------------+ | | | | | | | | | | +--------v----+ | | | | | | | Translation | | | | | | | +-------------+ | | | | | | | | Intelligent | | | | +-------------+ | | | | | | | Verification| | | Guarantee | | | | +-------------+ | | | | | |Intent Translation | | Module | | | | Module | | | | | +-------------------+ | | | | | | | | +-------------------+ | | | | |Intelligent Policy | | | | | | Mapping Module | | | | | +-------------------+ +-------------+ | | | +--------------------^-------------------+ | Southbound Interface +--------------------v-------------------+ | Infrastructure Layer | +----------------------------------------+ Among these layers, the intent-enabled layer is the core of this network management architecture. First, the intent translation module translates declarative intent (expressed in the form of natural language) into the network intent that can be recognized by the system (specific ). There may be an intent conflict problem when multiple intents are input simultaneously. Thus, the intent translation module executes intent verification and intent conflict resolution before the intent is issued (measurable). Second, the intelligent policy mapping module provides customized policies (achievable) for specific intents according to various requirements and evaluates the current policy by rewarding values. After that, in order to complete the policy configuration within the time-bound, the intent guarantee module is needed to execute feature extraction and location on the collected alarm information. Then the fault information is fed back to the intelligent policy mapping module. Based on the above design, on one hand, this architecture can achieve full lifecycle automated network management with humans out of the loop. On the other hand, it converts service requirements (intent) into network policies and provides self-adapted customized service with a full lifecycle verification. The functions of each module in intent-enabled layer are described below. Intent Translation Module Intent is expressed in the form of natural language, which is ambiguous and does not follow specific forms. Thus, the intent translation module translates the intent expressed in a natural language through bidirectional long short-term memory (Bi-LSTM) and morphological rules, and outputs the network’s understandable and regularized intent. Meanwhile, the intent translation module analyzes the accuracy and completeness of the translated intent through intent verification and conflict resolution, and continuously monitors whether there are conflicts. The intent translation module is the first step to realizing intent. In short, the translation module provides a bridge between the users and the network and is responsible for ensuring the integrity and realizability of the input intent. In addition, the result of the intent translation module is a critical foundation for the intelligent policy mapping module. Intelligent Policy Mapping Module The intelligent policy mapping module is the process of intent realization, in can consist of policy repository, fuzzy decision tree, and deep Q network (DQN). A large number of atomic policies can be stored in the policy repository, which can be established by the historical policy and administrator operation and maintenance experience. In particular, atomic policy refers to the smallest policy unit that can be executed but cannot be split again, such as some functional node configuration policies (routing selection, service provider node selection, etc.). The function of the fuzzy decision tree is to generate new atomic policies for the policy repository or to adjust the existing atomic policies. DQN is a combination of neural network and reinforcement learning, which are used to reorganize the atomic policies into a new policy that satisfies the current intent requirements. The function of the neural network is to calculate the configuration scores for state-action pairs and outputs all actions. Then the action with the highest configuration score is selected as the configuration action according to the Q-learning principle. Intent Guarantee Module The function of the intent guarantee module is to monitor the network in real time, collect and analyze the data of the faults that have occurred, and predict the faults that have not occurred in order to ensure the normal operation of the network. First, a fault information table, including fault type, location, quantity, and occurrence time, is established based on the collected network abnormal information. Second, a deep neural evolutionary network is used to repair faults and feed back the repair results to the intent translation module in real time. A deep neural evolution network can not only repair faults, but also ensure the implementation of intents when network faults occur.

The complexity of the future communication network dictates that achieving IBNM is a long-term goal, which needs to be completed step by step. Based on the autonomous network levels , we gives the definition of the levels of IBNM as five levels: manual network management (Level 0), intent assistance network management (Level 1), partial autonomous IBNM (Level 2), high autonomous IBNM (Level 3), and full autonomous IBNM (Level 4). IBNM transforms the traditional fully static network into a fully dynamic network, and details are as follows: Manual network management: This is a fully static phase, and the full life cycle of network management is done by the network administrator. The network is unable to provide differentiated services for users. Intent assistance network management: Intent is first introduced in this phase. Manual network management is still there; however, the IBN has been involved to handle partial tasks. Meanwhile, this level always requires administrators to configure parameters and correct for errors in implementation results. Partial autonomous IBNM: This level achieves a leap from static to dynamic. Most scenarios can achieve automatic configuration and timing verification for partial scenarios, which lessens the role of the administrator. In addition, the network can provide partial differentiated services for users at this micro-dynamic level. High autonomous IBNM: This level is semi-dynamic. The IBN can recognize multi- form (graph, text, and voice) intent input and realize autonomous intent translation. In addition, the IBN adopts real-time data collection and analysis methods, which improve the performance of network management in terms of intelligence, real time, and accuracy. Full autonomous IBNM: Full dynamicity is the objective of the IBN. The unique characteristics of a network management are intelligent intent insight, intelligent translation, intelligent configuration and error correction as well as real-time verification. All the tasks are done by the network. The IBN can provide users with differentiated services in full scenarios.

The advantages of the intent based network management include: A specific intent based network management architecture is proposed for sensing diversified service intents with various scenarios, objectives, and preferences. The proposed framework can achieve zero touch management within a time bound, which can handle the complex services in a fully automatic manner. The performance benefits are in terms of adaptivity and scalability, which rewards various network environments and avoids network reconstructions.

In order to evaluate the performance of the proposed IBNM architecture, we use the intent-based dynamic service function chain (SFC) as an example to solve the network management challenges (e.g., cross-domain orchestration and service functions are tightly coupled with the underlying equipment). At the same time, we developed an Openstack-based IBNM platform. The system demonstration implements the whole process from intent input to intent translation to intent policy generation to intent deployment, and the details are as follows. The user input cross-domain link-building requests (intent) in natural language at the web-page: Transfer a common-level video service from user A in Beijing to user B in Nanjing while constraining the execution time of the intent. The intent translation module outputs a conflict-free translation result, which indicates that the external input and the translation platform have been communicated. The translation results are intent tuples, which are displayed on the front-end interface in the form of name-value pairs. After the intent translation module, the translation results will be converted to JavaScript Object Notation (JSON) and transmitted to the intelligent policy mapping module. The intelligent policy mapping module divides the JSON request into an SFC: service function 1 (network address translation) service function 2 (firewall), and constructs the SFC request (name, tenant_id, description, service requirements, etc.). Then query whether there is an atomic policy combination that satisfies the current intent requirements in the policy repository. Following that, SFC is constructed based on the SFC interface, which is extended by Neutron. OpenStack schedules network resources, constructs sub-nets and ports, and generates two-dimensional space topology. Meanwhile, during the SFC construction process, the intent guarantee module monitors and manages network resource utilization as well as network failures in real time. Overall, IBNM achieves the decoupling of service application and network, and cross-domain network orchestration, while reducing the complexity of network management.

TBD

This document has no requests to IANA.