Compute-Aware RAN Resource Allocation via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Allocating resources in a radio access network (RAN) to maintain service-level guarantees across multiple slices is challenging, especially with the increasing demand for dynamic resource management and changing network conditions.
Innovation Solution
A network slice controller generates optimal RAN resource configurations using a machine learning model trained on data mapping compute resources needed at different stages of the physical layer software processing pipeline, allowing for dynamic allocation of resources among RAN slices based on service-level agreements and channel state information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional specialized hardware with fixed capacity is used to accommodate peak load, then service-level guarantees can be maintained, but resource utilization efficiency deteriorates during low-demand periods
Solution Approach 1:
The patent implements dynamic resource allocation in RAN by allowing resource configurations to change in real-time based on network conditions and service requirements. The system transitions from static hardware capacity to dynamic resource provisioning, where compute resources are allocated flexibly across different slices according to actual demand, thereby maintaining service-level guarantees while improving resource utilization efficiency during varying load conditions
Solution Approach 2:
The system changes key operational parameters including compute resource allocation, resource block allocation, and configuration parameters dynamically. By adjusting these parameters based on network conditions and service-level requirements, the system can optimize resource usage while maintaining reliability, moving away from fixed hardware capacity to adaptable parameter-based resource management
2Adaptability or versatility
If virtualization technologies are used to enable dynamic resource allocation, then resource flexibility improves, but difficulty in allocating resources to maintain service-level guarantees increases
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously monitors network conditions, service-level agreement compliance, and resource usage. This feedback is used to adjust resource allocations dynamically, providing the intelligence needed to manage complex virtualized environments effectively. The feedback loop enables automated decision-making that reduces the perceived complexity of resource allocation while maintaining service-level guarantees
Solution Approach 2:
The system introduces intermediary components including a resource configuration manager and machine learning models that act as mediators between virtualization infrastructure and service requirements. These intermediaries simplify the complex task of resource allocation by translating service-level requirements into appropriate resource configurations, thereby reducing allocation complexity while maintaining flexibility
3Reliability
If more compute resources are allocated to meet service-level agreements across multiple slices, then reliability improves, but computational power requirements increase
Solution Approach 1:
The patent applies local quality by allocating compute resources differently to different slices based on their specific service-level requirements. Instead of uniformly allocating resources across all slices, the system tailors resource allocation to each slice's needs, providing higher compute power to latency-sensitive slices and lower allocation to throughput-oriented slices, thereby meeting service-level agreements while minimizing total computational power requirements
Solution Approach 2:
The system implements partial resource allocation where compute resources are allocated only to the extent necessary to meet service-level guarantees for each slice. Rather than over-provisioning all slices with maximum resources, the system applies partial action by allocating just enough compute power to satisfy each slice's requirements, thereby reducing total computational power consumption while maintaining reliability
Data Source
AI summary
Aspects of the present disclosure relate to allocating RAN resources among RAN slices using a machine learning model. In examples, the machine learning model may determine an optimal RAN resource configuration based on compute power needs. As a result, RAN resource allocation generation and compute power requirements may improve, even in instances with changing or unknown network conditions. In examples, a prediction engine may receive communication parameters and/or requirements associated with service-level agreements (SLAs) for applications executing at least partially at a device in communication with the RAN. The RAN may generate one or more RAN resource configuration for implementation among RAN slices. Upon a change in network conditions or SLA requirements, an optimal RAN configuration may be determined in terms of required compute power.