Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AdaSlicing: Adaptive Online Network Slicing under Continual Network Dynamics in Open Radio Access Networks

Ming Zhao, Yuru Zhang, Qiang Liu, Ahan Kak, Nakjung Choi

TL;DR

This work tackles online network slicing under continual dynamics in Open RAN by proposing AdaSlicing, which combines per-slice Bayesian learning with an ADMM-based coordinator and a soft-isolated RAN virtualization layer. The system decomposes the global optimization into slice-level subproblems and a central coordination problem, solved via constrained Bayesian optimization and a convex coordination step, respectively, while enabling resource sharing through svRBs and SWs. Empirical evaluation on an end-to-end O-RAN testbed shows substantial gains: a $64.2\%$ reduction in total operating cost and a $45.5\%$ improvement in normalized slice performance, along with rapid convergence and strong adaptability to time-varying dynamics. The approach offers a practical, scalable path toward autonomous, fine-grained network slicing with performance guarantees in real-world Open RAN deployments.

Abstract

Open radio access networks (e.g., O-RAN) facilitate fine-grained control (e.g., near-RT RIC) in next-generation networks, necessitating advanced AI/ML techniques in handling online resource orchestration in real-time. However, existing approaches can hardly adapt to time-evolving network dynamics in network slicing, leading to significant online performance degradation. In this paper, we propose AdaSlicing, a new adaptive network slicing system, to online learn to orchestrate virtual resources while efficiently adapting to continual network dynamics. The AdaSlicing system includes a new soft-isolated RAN virtualization framework and a novel AdaOrch algorithm. We design the AdaOrch algorithm by integrating AI/ML techniques (i.e., Bayesian learning agents) and optimization methods (i.e., the ADMM coordinator). We design the soft-isolated RAN virtualization to improve the virtual resource utilization of slices while assuring the isolation among virtual resources at runtime. We implement AdaSlicing on an O-RAN compliant network testbed by using OpenAirInterface RAN, Open5GS Core, and FlexRIC near-RT RIC, with Ettus USRP B210 SDR. With extensive network experiments, we demonstrate that AdaSlicing substantially outperforms state-of-the-art works with 64.2% cost reduction and 45.5% normalized performance improvement, which verifies its high adaptability, scalability, and assurance.

AdaSlicing: Adaptive Online Network Slicing under Continual Network Dynamics in Open Radio Access Networks

TL;DR

This work tackles online network slicing under continual dynamics in Open RAN by proposing AdaSlicing, which combines per-slice Bayesian learning with an ADMM-based coordinator and a soft-isolated RAN virtualization layer. The system decomposes the global optimization into slice-level subproblems and a central coordination problem, solved via constrained Bayesian optimization and a convex coordination step, respectively, while enabling resource sharing through svRBs and SWs. Empirical evaluation on an end-to-end O-RAN testbed shows substantial gains: a reduction in total operating cost and a improvement in normalized slice performance, along with rapid convergence and strong adaptability to time-varying dynamics. The approach offers a practical, scalable path toward autonomous, fine-grained network slicing with performance guarantees in real-world Open RAN deployments.

Abstract

Open radio access networks (e.g., O-RAN) facilitate fine-grained control (e.g., near-RT RIC) in next-generation networks, necessitating advanced AI/ML techniques in handling online resource orchestration in real-time. However, existing approaches can hardly adapt to time-evolving network dynamics in network slicing, leading to significant online performance degradation. In this paper, we propose AdaSlicing, a new adaptive network slicing system, to online learn to orchestrate virtual resources while efficiently adapting to continual network dynamics. The AdaSlicing system includes a new soft-isolated RAN virtualization framework and a novel AdaOrch algorithm. We design the AdaOrch algorithm by integrating AI/ML techniques (i.e., Bayesian learning agents) and optimization methods (i.e., the ADMM coordinator). We design the soft-isolated RAN virtualization to improve the virtual resource utilization of slices while assuring the isolation among virtual resources at runtime. We implement AdaSlicing on an O-RAN compliant network testbed by using OpenAirInterface RAN, Open5GS Core, and FlexRIC near-RT RIC, with Ettus USRP B210 SDR. With extensive network experiments, we demonstrate that AdaSlicing substantially outperforms state-of-the-art works with 64.2% cost reduction and 45.5% normalized performance improvement, which verifies its high adaptability, scalability, and assurance.
Paper Structure (18 sections, 13 equations, 16 figures, 2 tables)

This paper contains 18 sections, 13 equations, 16 figures, 2 tables.

Table of Contents

  1. Introduction
  2. $AdaSlicing$ Overview
  3. System Model
  4. Proposed Solution
  5. Problem Decomposition
  6. The Design of Coordinator
  7. The Design of Learning Agents
  8. Soft-Isolated RAN Virtualization
  9. System Implementation
  10. Testbed Specifications
  11. Slice Applications
  12. Performance Evaluation
  13. Overall Performance
  14. AdaSlicing Dissection
  15. Soft-Isolated RAN Virtualization
  16. ...and 3 more sections

Figures (16)

  • Figure 1: The overview of AdaSlicing.
  • Figure 2: An example of runtime utilization of vRBs under different applications. Here, we run mixed applications before 400s, only watch live video in [400s, 600s], and then perform speedtest.
  • Figure 3: The architecture of soft-isolated RAN virtualization.
  • Figure 4: An example of soft-isolated RAN virtualization.
  • Figure 5: The overview of AdaSlicing testbed.
  • ...and 11 more figures