Queue-Aware vRAN Resource Scheduling Across Shared O-RAN Accelerators
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing O-RAN standard load balancing is inefficient as NFs cannot effectively share heterogeneous accelerating resources, leading to overloading of fastest hardware accelerators, poor performance, and increased energy consumption.
Innovation Solution
A method involving an access network controller that provides scheduling policies based on statistical information about queue waiting times and hardware accelerator processing capabilities, optimizing resource allocation to minimize energy consumption while meeting processing deadlines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If NFs greedily select fastest hardware accelerators without queue information, then processing speed is improved, but energy consumption increases and load balancing deteriorates
Solution Approach 1:
The system implements feedback mechanisms where the AAL-Broker collects queue waiting time statistics from all NFs and uses this information to make informed LPU assignment decisions. This feedback loop enables the system to balance load across heterogeneous accelerators while meeting processing deadlines, resolving the contradiction between speed and energy consumption.
Solution Approach 2:
The system performs preliminary analysis of queue waiting times and hardware accelerator capabilities before assigning LPUs to NFs. By predicting future queue states and pre-planning assignments, the system optimizes resource allocation to minimize energy consumption while ensuring processing deadlines are met.
2Speed
If NFs greedily select fastest hardware accelerators, then processing speed is improved, but load balancing deteriorates
Solution Approach 1:
The system dynamically adjusts LPU assignment strategies based on real-time queue waiting time statistics and hardware accelerator performance characteristics. This dynamic adaptation enables optimal load distribution across heterogeneous accelerators, improving both processing speed and load balancing efficiency simultaneously.
Solution Approach 2:
The system changes the assignment parameters from simple speed-based greedy selection to a multi-parameter approach considering queue waiting times, hardware capabilities, and historical performance data. This parameter transformation enables sophisticated load balancing while maintaining high processing speeds.
3Power
If heterogeneous accelerating resources are used, then processing capability is improved, but system complexity increases
Solution Approach 1:
The system segments the hardware accelerator resources into distinct LPUs, each associated with specific HAs. This segmentation allows independent management and optimization of each accelerator type while providing a unified interface through the AAL-Broker, reducing the complexity of managing heterogeneous resources.
Solution Approach 2:
The AAL-Broker acts as an intermediary layer between NFs and hardware accelerators, abstracting the complexity of heterogeneous resource management. This mediator collects statistics, makes informed assignments, and coordinates resource usage, simplifying the system architecture while enabling effective utilization of diverse accelerators.
Data Source
AI summary
A method performed by an access network controller is provided. The method includes providing, to a distributed unit (DU) of an access network, at least one scheduling policy for scheduling transmission of at least one transport block (TB) by a user equipment (UE). The at least one scheduling policy is based on statistical information about a respective waiting time, associated with each queue of a plurality of queues, each queue of the plurality of queues being associated with a respective logical processing unit (LPU) representing a hardware accelerator (HA) for processing TBs.