TL;DR
This paper introduces a novel QoS-aware scheduling algorithm for millimeter-wave (mmWave) wireless backhaul networks that leverages full-duplex (FD) communication to significantly boost the number of flows meeting their Quality of Service (QoS) requirements, especially in resource-constrained scenarios. By considering both residual self-interference (RSI) and multi-user interference (MUI), the proposed algorithm achieves higher system throughput and more satisfied flows compared to existing methods.
What Problem Does This Paper Address?
The rapid growth of 5G mobile networks, particularly the densification of small cells, has led to a massive increase in backhaul traffic. While mmWave bands offer high bandwidth, making wireless backhaul a cost-effective alternative to fiber, concurrent transmissions to maximize throughput introduce significant interference. Traditional scheduling algorithms often assume half-duplex (HD) communication, which limits simultaneous transmission and reception. Although full-duplex (FD) communication, enabled by self-interference (SI) cancellation technology, promises to double spectral efficiency, it introduces residual SI (RSI) alongside multi-user interference (MUI). Furthermore, many existing FD scheduling schemes assume ample time slot (TS) resources and don't specifically prioritize meeting the Quality of Service (QoS) requirements of individual flows, which is crucial for bandwidth-intensive applications like uncompressed video streaming. This paper addresses the challenge of efficiently scheduling FD mmWave wireless backhaul traffic in scenarios with limited TS resources, aiming to maximize the number of flows that satisfy their QoS requirements.
Methods
The core of this paper is the development of a QoS-aware Full-Duplex Concurrent Scheduling Algorithm. The approach involves:
- Leveraging Full-Duplex (FD) Communication: The algorithm explicitly incorporates the ability of base stations (BSs) to transmit and receive simultaneously, exploiting the potential for increased spectral efficiency.
- Contention Graph Construction: A contention graph is built where each vertex represents a flow. An edge is added between two flows if they cannot be concurrently scheduled. This inability arises from two main factors:
- FD Constraints: Two flows cannot be scheduled concurrently if they both use the same BS as a transmitter or both use it as a receiver.
- Interference Threshold: If the relative interference (RI) between two flows exceeds a predefined contention threshold ($\sigma$), an edge is added, indicating they cannot be scheduled together. The RI is calculated considering both MUI and RSI.
- QoS-Aware Scheduling: The algorithm prioritizes flows based on their estimated time to completion ($\xi_f$) if they were to transmit without interference. Flows that are unlikely to complete within the available time slots (M) are removed to reduce complexity and focus on more achievable flows.
- Iterative Scheduling: The algorithm iteratively schedules flows slot by slot. In each time slot, it attempts to schedule the flow with the smallest $\xi_f$ that has no contention with ongoing transmissions and whose scheduling would improve overall system throughput. A flag
changetracks if any flow has completed its QoS requirement in the current slot, triggering a re-evaluation of scheduling decisions. - Optimization Formulation: The problem is formulated as a nonlinear integer programming (NLIP) problem aiming to maximize the number of completed flows, subject to constraints on BS capacity (max two flows per BS) and the FD transmission rule.
Datasets
Not discussed in the paper. The evaluation is based on extensive simulations with parameters representative of a 60GHz mmWave wireless backhaul network.
Key Insights and Ideas
- FD is a Game-Changer for Resource-Constrained Backhaul: The paper demonstrates that effectively utilizing FD communication, even with residual SI, can significantly outperform HD schemes when TS resources are limited.
- QoS is Paramount: Simply maximizing throughput isn't enough. Prioritizing flows based on their QoS requirements and estimated completion time is crucial for achieving a higher number of satisfied users.
- Interference Management is Key: The algorithm's success hinges on accurately modeling and managing both MUI and RSI. The contention graph, incorporating a threshold for relative interference, is a critical tool for this.
- The Trade-off with SI Cancellation: The performance of FD schemes is highly sensitive to the level of SI cancellation. Higher cancellation levels (smaller $\beta$) lead to better performance, highlighting the ongoing importance of SI cancellation research.
- Contention Threshold Tuning: The choice of the contention threshold ($\sigma$) is critical. Too low a threshold restricts concurrent transmissions, while too high a threshold can lead to excessive interference, negatively impacting QoS.
Key Results
The proposed QoS-aware FD concurrent scheduling algorithm (Proposed-FD) consistently outperforms existing schemes:
- Increased Completed Flows: The Proposed-FD algorithm significantly increases the number of flows that satisfy their QoS requirements, especially as the number of flows increases. For instance, with 90 flows, it improves completed flows by 29.9% compared to the best HD baseline (MQIS) and 30.1% compared to its HD FD counterpart (Proposed-HD).
- Higher System Throughput: The algorithm also achieves higher system throughput. Compared to FDP (a non-QoS-aware FD scheme), Proposed-FD improves system throughput by 34.1% (with 90 flows).
- Impact of SI Cancellation: Performance degrades as SI cancellation levels decrease ( $\beta$ increases). When $\beta$ becomes very large, the performance of Proposed-FD approaches that of Proposed-HD, indicating that poor SI cancellation negates the benefits of FD.
- Impact of Contention Threshold: The algorithm's performance is sensitive to the contention threshold ($\sigma$). An optimal threshold (around $10^{-3}$ in simulations) balances concurrent transmissions and interference management, leading to the best results.
Limitations and Future Work
- Limitations:
- The paper acknowledges that the performance of FD schemes is highly dependent on the SI cancellation level. If SI cancellation is poor, the benefits of FD are diminished.
- The optimal contention threshold needs careful tuning for best performance.
- Future Work:
- The authors suggest incorporating the problem of blockage in mmWave communications into future research, proposing a robust scheme for mmWave FD backhaul networks.
Summary updated