Multi-Connectivity for Multicast Video Streaming in Cellular Networks

Sadaf ul Zuhra, Prasanna Chaporkar, Abhay Karandikar, H. Vincent PoorNetwork 2024 · DOI

TL;DR

This paper tackles the challenge of delivering high-quality video streams in cellular networks by proposing and analyzing a "Multi-Connectivity Multicast" (MC Multicast) system. By allowing users to connect to multiple base stations simultaneously for multicast streams, the system significantly improves user coverage and resource utilization, especially for cell-edge users.

What Problem Does This Paper Address?

The explosion of video streaming, particularly live events, is straining cellular network resources. Traditional unicast (one-to-one) streaming is inefficient for mass audiences, consuming excessive bandwidth. Multicast (one-to-many) streaming offers a solution by sharing content over common radio resources. However, multicast performance can be bottlenecked by users with poor signal quality (cell-edge users). Multi-connectivity (MC), a feature allowing devices to connect to multiple base stations simultaneously, has shown promise for improving individual user connectivity. This paper investigates the synergistic benefits of combining MC with multicast streaming to enhance video delivery efficiency and user experience in cellular networks.

Methods

The authors propose an MC Multicast system that integrates with existing 3GPP Multimedia Broadcast Multicast Services (MBMS). The core of their work lies in addressing the resource allocation problem for this system, aiming to maximize the number of simultaneously served users.

  • Resource Allocation Problem Formulation: They define the problem as selecting the optimal set of radio resource blocks (PRBs) across multiple cells to serve the maximum number of users.
  • NP-Hardness Proof: They prove that this resource allocation problem is NP-hard, meaning no efficient (polynomial-time) algorithm can guarantee an optimal solution for large systems. This is achieved by reducing it to the known NP-hard Maximum Coverage Problem (MCP).
  • Centralized Greedy Approximation Algorithm (CGA): To overcome the NP-hardness, they develop a centralized greedy algorithm. This algorithm iteratively selects PRBs that serve the most new users until all cells have a PRB allocated. They prove this algorithm achieves an approximation factor of (1 - 1/e), meaning it guarantees a solution that is at least (1 - 1/e) times the optimal.
  • Distributed Greedy Allocation (DGA): Recognizing that centralized control can be a bottleneck, they also propose a distributed algorithm. In DGA, each base station independently allocates its PRBs to maximize users within its own cell, sacrificing some optimality for lower latency and reduced signaling overhead.

Datasets

The authors utilized traces from actual video streams to generate realistic traffic patterns for their simulations. Specifically, they mention using traces from:

  • A video of the Tokyo Olympics (obtained from http://trace.eas.asu.edu/).

Key Insights and Ideas

  • Synergy of MC and Multicast: The combination of multi-connectivity and multicast streaming is not just additive; it creates a synergistic effect that significantly boosts performance, particularly for users at the cell edge.
  • MC Multicast as a Simpler MBSFN Alternative: MC Multicast offers similar benefits to MBSFN (Multi-Base Station Forward Network) in terms of improved signal quality for cell-edge users but with a simpler framework and reduced synchronization overhead.
  • The Power of Greedy Approximation: Even though the optimal resource allocation is NP-hard, a well-designed greedy approach can provide a strong approximation guarantee, making it practical for real-world deployment.
  • Trade-off Between Centralization and Distribution: While centralized algorithms (like CGA) offer better performance, distributed algorithms (like DGA) are crucial for scenarios where centralized control is infeasible due to latency or overhead, and they still provide competitive results.
  • Significant User Uplift: The proposed MC Multicast system can serve a substantial portion of users who would otherwise be left unserved in a single-connectivity multicast scenario.

Key Results

  • Improved User Coverage: MC Multicast successfully serves approximately 60% more users compared to single-connectivity multicast using the same resources.
  • Approximation Guarantee: The proposed Centralized Greedy Approximation (CGA) algorithm achieves a (1 - 1/e) approximation factor for the resource allocation problem.
  • Distributed Performance: The Distributed Greedy Allocation (DGA) algorithm, despite being distributed, incurs only a marginal increase (at most 0.3%) in packet loss compared to the centralized CGA algorithm.
  • Enhanced Packet Delivery: Simulations show that MC Multicast delivers significantly more packets successfully than Single-Connectivity (SC) multicast, especially as the number of users increases. For instance, MC Multicast delivered around 8000 more packets than SC Multicast in one simulation scenario.
  • Performance Gap: The performance gap between MC and SC multicasting widens as the number of users and cell radius increase, highlighting the benefits of MC in larger and more challenging scenarios.
  • Outperforms MBSFN: MC Multicasting significantly outperforms MBSFN in terms of packet delivery and user serving capacity, while also eliminating the need for strict synchronization.

Limitations and Future Work

  • Static User Assumption: The current work assumes users are static throughout the multicast transmission. The impact of user mobility on the proposed algorithms needs to be studied, requiring new resource allocation algorithms that can adapt to changing user sets.
  • Machine Learning for Optimization: Given the NP-hard nature of the problem, machine learning-based algorithms could be explored for optimizing allocation decisions.
  • Multiple Multicast Streams: The research could be extended to systems supporting multiple simultaneous multicast streams within a region.

Summary updated

BibTeX
@article{zuhra2024multi,
  author = {Sadaf ul Zuhra and Prasanna Chaporkar and Abhay Karandikar and H. Vincent Poor},
  title = {{Multi-Connectivity for Multicast Video Streaming in Cellular Networks}},
  journal = {Network},
  year = {2024},
  doi = {10.3390/network4020009},
}
Figure