RAN Slicing in Multi-MVNO Environment Under Dynamic Channel Conditions

Thomas Hou, Darshan A. Ravi, Vijay K. Shah, Chengzhang Li, Y. Thomas Hou 0001, Jeffrey H. ReedIEEE Internet of Things Journal 2022 · DOI

TL;DR

This paper tackles the challenge of efficiently allocating radio access network (RAN) resources to multiple Mobile Virtual Network Operators (MVNOs) in a 5G environment, considering that channel conditions change dynamically. They propose a novel MCS-aware RAN Slicing (MaRS) algorithm that aims to maximize the number of served users while meeting their Quality of Service (QoS) requirements, demonstrating near-optimal performance compared to a theoretical upper bound.

What Problem Does This Paper Address?

The advent of 5G and the Internet of Things (IoT) has led to an explosion of diverse wireless services with varying Quality of Service (QoS) demands. To manage this complexity, RAN slicing has emerged as a crucial technology, allowing network operators (NOs) to partition their physical infrastructure into logical segments, each tailored to specific MVNO requirements. However, efficiently allocating scarce radio resources (like resource blocks or RBs) to these slices, especially when channel conditions fluctuate, is a significant challenge. Existing approaches often fall short by not adequately considering dynamic channel conditions or by offering limited flexibility to MVNOs in scheduling their users. This paper addresses the problem of MCS-aware RAN slicing (MaRS) in a multi-MVNO environment under dynamic channel conditions, aiming to maximize the number of users served while ensuring their minimum data rate requirements are met.

Methods

The authors formulate the MaRS problem as a Non-Linear Programming (NLP) problem, which they prove to be NP-Hard. To address this, they propose a greedy algorithm-based solution. The core of their approach involves:

  • MCS-Awareness: The algorithm explicitly considers the Modulation and Coding Scheme (MCS) level, which directly impacts the achievable data rate on a given resource block and is influenced by the Channel Quality Indicator (CQI) reported by users.
  • Dynamic Channel Conditions: The algorithm accounts for the fact that channel conditions can change over time, meaning the optimal MCS for a resource block can vary.
  • Greedy Allocation: The algorithm iteratively allocates resource blocks. It prioritizes users based on a combined sorting of their minimum data rate requirements and their MVNO-defined scheduling order. For each user, it greedily selects resource blocks with the highest possible MCS that still meets the user's data rate requirement, thereby minimizing RB utilization.
  • MVNO Flexibility: The proposed architecture allows each MVNO to implement its own scheduling algorithm, providing flexibility.

Datasets

Not discussed in the paper. The authors simulate network scenarios and generate channel conditions rather than using pre-existing datasets.

Key Insights and Ideas

  • MCS is Key for Efficiency: Choosing a higher MCS level, when channel conditions permit, significantly reduces the number of resource blocks needed to meet a user's data rate requirement. This is a critical insight for maximizing the number of served users.
  • Greedy Approach Works Well: Despite the NP-hard nature of the problem, a well-designed greedy algorithm, prioritizing users based on a combination of QoS and MVNO scheduling, can achieve near-optimal performance.
  • MVNO Flexibility is Valuable: Allowing MVNOs to define their own scheduling order provides a crucial layer of customization and control, which is a key differentiator from some prior work.
  • Dynamic Conditions are Critical: Ignoring dynamic channel conditions can lead to suboptimal resource allocation and a reduced ability to serve users.

Key Results

The paper presents simulation results demonstrating the effectiveness of their MaRS algorithm:

  • Near-Optimal Performance: The MaRS algorithm consistently achieves performance within 5-10% of a theoretical upper bound across various simulation scenarios, including different Rician factors (representing channel fading), varying time slot durations, and different network configurations.
  • RB Utilization: The MaRS algorithm, by employing the Maximum MCS selection criteria, significantly outperforms static allocation algorithms (Average MCS, Lowest MCS) in terms of RB utilization, meaning it serves more users with fewer resources.
  • Scalability: The algorithm is shown to be scalable and can be applied to non-real-time and near-real-time RIC deployments in O-RAN architectures.
  • Fast-Changing Channels: Even under fast-changing channel conditions (where conditions vary at each Transmission Time Interval - TTI), the MaRS algorithm maintains performance close to the upper bound.

Limitations and Future Work

  • Upper Bound as Benchmark: The paper uses a theoretical upper bound as a benchmark, which assumes ideal conditions (excellent channel, no correlation). While useful, real-world performance might still have a gap.
  • MVNO Throughput Limit: The paper acknowledges that setting the maximum throughput limit per MVNO slice (Am) is a business decision and its optimal selection is outside the scope of this work.
  • Future Work: The authors suggest exploring more advanced scheduling algorithms and investigating the impact of different MVNO business models on resource allocation. They also mention the potential for further optimization by considering inter-cell interference and more complex network topologies.

Summary updated

BibTeX
@article{hou2022ran,
  author = {Thomas Hou and Darshan A. Ravi and Vijay K. Shah and Chengzhang Li and Y. Thomas Hou and Jeffrey H. Reed},
  title = {{RAN Slicing in Multi-MVNO Environment Under Dynamic Channel Conditions}},
  journal = {IEEE Internet of Things Journal},
  year = {2022},
  doi = {10.1109/jiot.2021.3108145},
}
Figure