Abstract
This paper investigates the scalability of content delivery in fading multi-input single-output (MISO) channels with cache-aided users. As the number of users ($K$) grows, maintaining high throughput becomes challenging. The authors demonstrate that by combining coded caching with wireless multicasting, linear scaling of the content delivery rate—where total throughput grows proportionally with $K$—is achievable even without channel state information at the transmitter (CSIT). This is possible if the number of transmit antennas or sub-channels scales logarithmically with $K$. Additionally, a threshold-based user selection scheme requiring only one-bit feedback is shown to maintain this scalability.
The Challenge of Wireless Scalability
In conventional wireless networks, increasing the user count typically degrades individual performance as resources are split. Two primary technologies aim to solve this: Massive MIMO, which uses large antenna arrays for spatial multiplexing, and Coded Caching, which pre-stores content to enable efficient multicasting.
However, Massive MIMO often requires precise Channel State Information at the Transmitter (CSIT), and standard coded caching models frequently ignore the impact of wireless fading, where a single user's poor connection can bottleneck an entire multicast group. This research explores how to robustly combine these technologies in realistic fading environments to ensure the network remains truly scalable.
Methodology: Achieving Linear Scaling
The authors analyze a system with an $n_t$-antenna base station serving $K$ users through several innovative approaches:
Threshold-based User Selection
To prevent "bottleneck" users from slowing down the network, the researchers propose a scheme where users provide a single bit of feedback. This bit indicates whether their current signal-to-noise ratio (SNR) exceeds a specific threshold. The base station then only serves users with "good" channels, maintaining high-speed delivery without complex feedback overhead.
Mixed Delivery Strategy
When CSIT is available, a "rate-splitting" framework is used. This strategy simultaneously transmits a common multicast signal (leveraging coded caching) and private unicast signals (leveraging spatial multiplexing). The power allocation between these two streams is optimized to maximize total throughput, particularly when user memory is scarce.
Key Insights and Results
The study provides several critical theoretical and numerical findings:
- The Logarithmic Rule: Linear scaling does not require a linear increase in hardware. If the product of transmit antennas ($n_t$) and independent sub-channels ($L$) scales as $\ln(K)$, the system remains scalable without CSIT.
- Channel Hardening: In large antenna arrays, channel quality tends to stabilize. This "hardening" allows multicasting to remain efficient even for massive user groups.
- Linear Throughput Growth: Numerical simulations confirm that the equivalent content delivery rate grows linearly with the number of users under the proposed conditions, significantly outperforming systems without scheduling or proper antenna scaling.
Conclusion
The research proves that wireless networks can achieve linear throughput scaling by intelligently combining multi-antenna multicasting with coded caching. While the current model assumes i.i.d. Rayleigh fading and static users, it lays the groundwork for future studies on non-homogeneous path loss, user mobility, and fairness constraints in massive cache-aided networks.
Summary updated