TL;DR
This paper introduces a novel Quality of Service (QoS) MAC scheduler designed for eXtended Reality (XR) traffic in 5G New Radio (NR) networks. By leveraging the 5G NR QoS flow model, the proposed scheduler effectively classifies and prioritizes mixed XR traffic based on assigned QoS indicators, demonstrating improved performance in both simulated and realistic network conditions.
What Problem Does This Paper Address?
eXtended Reality (XR) applications, encompassing Virtual Reality (VR), Augmented Reality (AR), and Cloud Gaming (CG), are rapidly gaining traction and are poised to become a major computing platform. These applications demand stringent Quality of Service (QoS) in terms of low latency and high throughput to deliver an immersive user experience. However, existing 5G NR networks face challenges in efficiently managing the diverse and often conflicting QoS requirements of these mixed XR traffic flows. The 5G NR QoS model, with its concept of QoS flows, offers a promising framework, but effective scheduling mechanisms are needed to fully exploit its capabilities for XR. This paper addresses the need for a specialized QoS MAC scheduler that can intelligently differentiate and prioritize various XR traffic types based on their unique QoS requirements.
Methods
The core contribution of this paper is the design and evaluation of a new QoS MAC scheduler. This scheduler operates by calculating a "scheduling weight" for each Data Radio Bearer (DRB), which represents its priority for being scheduled. This weight is derived from a formula that considers several QoS indicators, including:
- Priority Level (P): A fundamental QoS attribute indicating the importance of the flow.
- Instantaneous Achievable Data Rate (r): The current data rate based on spectrum efficiency and channel bandwidth.
- Past Average Data Rate (R(τ)): A smoothed average of the data rate over a defined window.
- Retransmission Status (F): A factor that increases priority if a DRB has retransmission data.
- Delay Budget Factor (D): A novel factor introduced by the authors, which accounts for the delay-aware weight related to head-of-line (HOL) packet delay and the Packet Delay Budget (PDB). This factor is crucial for differentiating XR traffic with its stringent latency requirements.
The scheduler then orders users based on these descending weights, ensuring that higher-priority traffic is served first. The paper also details the implementation of this scheduler within the ns-3 5G-LENA simulator, extending it to support both TDMA and OFDMA access modes and incorporating XR traffic models.
Datasets
The paper does not introduce new datasets. Instead, it utilizes the ns-3 5G-LENA simulator, which has been previously calibrated using 3GPP reference scenarios, specifically the Rural-eMBB and Dense-eMBB configurations. These configurations are based on ITU-R recommendations for network modeling and serve as the basis for the evaluation scenarios.
Key Insights and Ideas
- The Importance of a Dedicated Delay-Aware Factor: The introduction of the "Delay Budget Factor (D)" is a key innovation. This factor directly addresses the critical low-latency requirements of XR applications, allowing the scheduler to proactively prioritize time-sensitive traffic.
- Prioritization in Saturated Networks: The scheduler demonstrates its ability to effectively prioritize traffic even when network resources are scarce (saturated conditions). This is crucial for real-world deployments where bandwidth limitations are common.
- Granular QoS Control: The scheduler's ability to differentiate traffic based on specific QoS indicators (like 5QI values) allows for fine-grained control over how different XR applications are treated, ensuring a better overall user experience.
- Validation Across Scenarios: The paper validates the scheduler's performance in both a simple, controlled scenario and a more complex, realistic multi-cell environment, providing strong evidence of its robustness.
Key Results
The paper presents simulation results demonstrating the effectiveness of the proposed QoS MAC scheduler:
- Non-Saturated Scenario: In conditions with ample bandwidth, all schedulers (proposed, RR, PF) performed similarly. However, when VR traffic was assigned a higher priority 5QI (5QI=87), the proposed scheduler clearly prioritized it, showing a slight increase in delay for lower-priority traffic. This confirms its ability to manage traffic based on priority.
- Saturated Scenario: When bandwidth was reduced to 75 MHz, the proposed scheduler continued to demonstrate its ability to prioritize VR traffic, leading to significant packet delay reduction and throughput improvement for VR, with only a minor impact on other traffic types.
- Multi-Cell Scenario: In a realistic multi-cell environment, the proposed scheduler effectively prioritized VR traffic in the downlink direction, even under high interference and saturated conditions, showcasing its capability to classify traffic based on QoS indicators. The uplink performance was found to be similar to traditional schedulers, as the uplink was not saturated.
Limitations and Future Work
- Limitations: The paper acknowledges that the implementation details of the XR traffic models are out of scope. While the scheduler was evaluated for both 1x1 and Mx1 mapping architectures, a deeper comparison of these architectures in the context of multi-QoS flows per user is suggested.
- Future Work: The authors envision extending the scheduler to consider additional QoS indicators, such as resource type (GBR/delay-critical GBR/non-GBR) and packet delay budget. They also suggest exploring further design options for calculating the scheduling weight in scenarios with multiple QoS flows per user, which would enable more sophisticated flow differentiation.
Summary updated