Multi-Cloud Software Fabric for Autonomous Network Connectivity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for wireless networks are not robust enough to meet the requirements of 5G networks, particularly in terms of seamless integration with hybrid and multi-cloud environments, fault containment, and quality-of-service, leading to challenges in providing scalable and cost-effective solutions for enterprises and operators.
Innovation Solution
A machine-learning based distributed, hybrid, and multi-cloud software fabric that unifies communication infrastructure across hybrid and multi-clouds, enabling seamless connectivity between small independent networks and public networks, and providing autonomous resource allocation and traffic management using AI/ML models for control and data layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems are hosted locally or on a single cloud provider, then system simplicity is maintained, but seamless enterprise integration with operator networks is not achieved
Solution Approach 1:
The system is divided into distinct layers (discovery layer, control layer, data layer) that can operate independently yet coordinate together. Each layer handles specific functions, allowing the system to integrate with multiple networks while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The discovery layer acts as an intermediary between client services and the distributed control layer functions. It receives connection requests, determines service requirements, and orchestrates resource allocation across multiple cloud providers, enabling seamless integration without direct complex point-to-point connections.
2Speed
If network functionality is centralized, then system management is simplified, but low-latency services and edge computing requirements are not met
Solution Approach 1:
The system places control layer functions at multiple distributed locations across different cloud providers and edge infrastructure. Each location provides services locally to reduce latency, while the discovery layer coordinates between them. This allows services to be delivered from the nearest available resource, minimizing transmission delays.
Solution Approach 2:
The system dynamically selects and allocates control layer functions based on real-time service requirements, resource availability, and performance metrics. Machine learning models continuously adapt resource allocation decisions to optimize latency and performance, allowing the system to respond flexibly to changing conditions rather than following static centralized routing.
3Reliability
If cloud infrastructure is built for scalability, then system expansion is enabled, but fault containment and robustness for 5G services are not achieved
Solution Approach 1:
The discovery layer, control layer, and data layer are segmented into independent functional units distributed across multiple cloud providers. This segmentation isolates faults to specific layers or providers, preventing cascading failures. If one provider experiences issues, other providers can continue serving requests without being affected.
Solution Approach 2:
The system uses machine learning models to dynamically adjust operational parameters such as resource allocation, load distribution, and failover thresholds based on real-time system state. When faults are detected, the models automatically change parameters to contain the fault and maintain service reliability, adapting the system behavior to current conditions.
4Productivity
If multiple machine-learning models are used for resource allocation, then service optimization is improved, but computational overhead and system complexity increase
Solution Approach 1:
Different machine learning models are assigned to different layers (discovery, control, data) based on their specific functions and requirements. Each model operates independently within its layer, making decisions relevant to that layer's resources and services. This segmentation allows parallel processing and reduces the complexity of managing a single monolithic model.
Solution Approach 2:
The discovery layer serves as an intermediary that coordinates between the client services and the control layer models. It pre-processes service requests, determines initial resource requirements, and passes this information to the control layer models, reducing the computational burden on downstream models and optimizing overall resource allocation efficiency.
Data Source
AI summary
The present disclosure describes an artificial intelligence (AI)/machine learning (ML) based distributed, hybrid, and multi-cloud software fabric-based system that unifies the communication infrastructure across hybrid and multi clouds. This mobile connectivity software fabric allows operators to modernize their networks to bring significant operational savings while rolling out new mobile services. This fabric can enable small independent networks and allow them to seamlessly connect with public networks, and it can enable network of networks while keeping the underlying compute and heterogeneity unified.