Network Data Store creates a data lake that stores different kinds of network data, which are further used for visualization, analytics, machine learning, and autonomous networks.

• Structured Data Store - The structured data store includes structured network data like inventory data, NE configuration data, planning data, and NE alarm data. The data store keeps the latest values and tracks changes over a period of time as well.
• Columnar Data Store - The columnar data store includes a column-based database suitable for large-volume data warehouse requirements. This data store keeps NE performance data and NE event data, supporting real-time analytics on large volumes of data.
• Vector Data Store – Vector data enables the development and deployment of Generative AI applications by converting unstructured text, images, and video data into vectors for fast semantic search and retrieval operations. Some examples of unstructured data include ticket data with recommended solutions, product documentation, acceptance report images, etc.

Network rules engine enables the creation of rules to validate that network KPIs are aligned with prescribed parameters. The rule language depends on the type of network data, e.g., inventory, configuration, event, alarms, or performance data.

• Inventory/Config Rules - These rules ensure that network inventory and configuration data are aligned with prescribed behavior. For example, auditing that the radio unit's configured frequency and power parameters match the planning data.
• Performance Rules - These rules enable extracting the health of network elements using periodic PM data. Rules can perform aggregation, bucketization, and logical and mathematical operations on different PM counters to extract network performance. For example, bucketizing NEs with peak-hour utilization over 90%.
• Alarms/Event Rules - These rules enable mapping of periodic NE events and alarms data to the health and performance of the network.

The Network Dashboard is the visualization layer of the autonomous network solution that depicts the results of the execution of network rules on the network data store. This enables a consolidated view of network health as well as a 360-degree insight into the health of individual NEs.

• Regional Network Dashboard - The key functionality of the network dashboard is to present a consolidated view of network health for a region.
• Device Dashboard - The device dashboard provides a 360-degree view of the device, including its current configuration, recent configuration changes, performance data trends, alarms, and events.
• Network Health Trends - Trends help visualize and track network and device health over a period of time, allowing for the gauging of any deterioration or improvement in network health KPIs.

The Intent Manager connects human intent to the network infrastructure. Intent abstracts expectations from the network in natural language, which need to be interpreted, translated, stored, and transported for network actions.

• Intent Interpreter - It analyzes natural language inputs from network operators, business stakeholders, or even customers to understand their intent (e.g., “optimize for latency” or “expand capacity”). It interprets human input and maps it to network tasks.
• Intent Translator - Once the intent is interpreted, generative models help autonomously translate these intents into network actions. This step involves generating policies and rules that result in actions/commands for specific network elements.
• Intent Monitor - It is responsible for continuously monitoring and analyzing ingested network data and notifying of deviations from the network intent. It also proactively identifies future deviations from intent based on trends in the data.

The Network Brain is the AI-based decision engine of the autonomous network. It uses various machine learning (ML) approaches, such as supervised learning, reinforcement learning, and generative AI, to analyze network data and expected intents and to make decisions on optimal actions.

• Decision Generator - The ML inference component of various algorithms generates a prioritized list of decisions based on intent monitoring of incoming network data and expected intents.
• Decision Validator – It validates decisions either by using a network digital twin or predefined rules and logic.
• Decision Executor - Using various domain and component orchestrators, it executes decisions that result in component-specific physical or configuration changes.

Autonomous networks strive to create a Dark NOC where network operations are completely automated. With intelligent technical support, technicians would be assisted by an AI advisor that would:

• Enterprise RAG Engine - This includes components for building an enterprise Retrieval Augmented Generation (RAG) pipeline, such as chunking, retrieval, reranker, guardrails, etc. This enables the ingestion and retrieval of various types of data, such as operating manuals, historical tickets, and root cause analysis documents.
• Enterprise Data Connectors – These include connectors that pull data from various sources, such as FTP servers, S3 repositories, Google Drive, NFS mount points, and more.
• AI Incident Advisor - It interacts with technical support personnel in natural language and provides recommended resolutions, including operating procedures. It can also automate many resolutions.