Abstract
Cloud Radio Access Networks (C-RAN) promise superior interference management by centralizing baseband processing at a Control Unit (CU). However, this centralization introduces "fronthaul latency," which can significantly impact the performance of Layer 2 Automatic Retransmission reQuest (ARQ) protocols. This paper utilizes a multi-dimensional Markov modeling approach to quantify the trade-offs between interference cancellation and network lag. Our analysis compares conventional, pure C-RAN, and hybrid architectures, demonstrating that while latency challenges simple protocols like Stop-and-Wait, advanced protocols like Selective Repeat and hybrid system designs can effectively harness the benefits of the cloud.
Introduction
In traditional cellular architectures, baseband processing occurs locally at the Base Station (BS). While this minimizes latency, it limits the ability to manage inter-cell interference. C-RAN architectures address this by migrating processing to a centralized Control Unit (CU), enabling joint processing of signals from multiple base stations to cancel interference.
The primary challenge in C-RAN is the "fronthaul" link connecting the BS to the CU. The time required to transmit data to the CU and receive feedback (ACK/NAK) introduces latency. For ARQ protocols, which rely on rapid feedback to manage packet retransmissions, this delay can lead to idle periods and reduced throughput. This study seeks to determine the conditions under which the interference-management gains of C-RAN outweigh the penalties of fronthaul delay.
Methodology
The authors model the system using a multi-dimensional Markov process to capture the dynamics of the wireless environment and the communication protocols. The framework is decomposed into three interacting sub-chains:
- Channel Sub-Chain: Models time-correlated Rayleigh fading to represent signal strength fluctuations.
- Interference Sub-Chain: A binary Markov model representing the presence or absence of external interference.
- Protocol Sub-Chain: Tracks the mobile terminal's buffer state and ARQ logic.
The study evaluates three architectures:
- Conventional: Local decoding at the BS (low latency, high interference).
- C-RAN: Centralized decoding at the CU (high latency, low interference).
- Hybrid C-RAN: A two-stage approach where the BS attempts local decoding first, forwarding to the CU only upon failure.
Key Insights and Results
The Hybrid Advantage
The Hybrid C-RAN architecture emerges as the most robust solution. At high signal-to-interference ratios (SIR), it functions like a conventional system, avoiding unnecessary cloud latency. At low SIR, it utilizes the CU's processing power to resolve interference, outperforming both standalone models.
Protocol Resilience to Latency
The choice of ARQ protocol is critical in high-latency environments. Stop-and-Wait (SW) performs poorly because it remains idle while waiting for feedback. In contrast, Selective Repeat (SR) and Go-Back-N (GBN) utilize transmission windows to continue sending data, making them far more suitable for C-RAN. For SR, C-RAN remains superior to conventional systems provided the fronthaul delay ($\Delta$) does not exceed 6 time slots.
Efficiency vs. Throughput
While latency may reduce raw throughput, C-RAN often improves overall transmission efficiency. By successfully canceling interference, the CU reduces the total number of required transmissions per packet, which can significantly extend mobile terminal battery life.
Conclusion and Future Work
This research highlights that the viability of C-RAN is deeply tied to the underlying ARQ protocol and the specific architecture employed. While fronthaul latency is a significant bottleneck, hybrid systems and window-based protocols like Selective Repeat can mitigate these effects. Future research should expand this model to include multi-user detection scenarios and more complex, multi-level interference models to better reflect real-world network conditions.
Summary updated