Abstract
Wireless Mesh Networks (WMNs) often struggle with the trade-off between reliability and latency, particularly for real-time video streaming. While link-layer retransmissions are designed to mitigate wireless noise, their impact in multi-hop environments is complex. This paper investigates how varying retransmission limits affect video quality (PSNR), delay, and jitter across different network loads and topologies. Our findings suggest that while a minimal number of retries significantly boosts quality, excessive retransmissions can lead to congestion collapse and unusable latency in multi-hop paths.
Introduction
The IEEE 802.11 standard, the backbone of most WMNs, employs a link-layer retransmission mechanism to handle packet loss. In single-hop scenarios, this is highly effective. However, in multi-hop mesh networks, each retry consumes shared airtime, potentially causing intra-flow and inter-flow interference. For delay-sensitive applications like video streaming, these compounded delays and increased congestion can severely degrade the user experience. The authors quantify these impacts using Peak Signal-to-Noise Ratio (PSNR) and network health metrics across various traffic loads and multi-hop topologies.
Experimental Setup
The researchers conducted experiments on a real-world indoor testbed to capture the nuances of wireless propagation. The setup involved a multi-hop mesh network using Soekris board Access Points and HP laptops equipped with Atheros-based wireless cards.
The software environment utilized the MadWifi driver on Linux, modified to allow granular control over the link-layer retransmission limit. Video was streamed using UDP/RTP to isolate MAC-layer effects from transport-layer congestion control. The test video was a 2000-frame "highway" sequence encoded in MPEG4 at 25 fps, tested at bitrates ranging from 500 Kbps to 2500 Kbps to simulate varying network loads relative to path capacity.
Results and Analysis
The study revealed a clear "law of diminishing returns" regarding retransmission limits. The most significant gain in video quality occurs when moving from zero to one retransmission.
As shown in Figure 6(a), PSNR tends to plateau after one or two retries for stable links (e.g., 2-hop 1000 Kbps). However, for high-bitrate streams operating near network capacity, especially in multi-hop scenarios (e.g., 3-hop and 4-hop 2500 Kbps), the PSNR remains significantly lower, indicating that retransmissions cannot overcome fundamental capacity limits. Once the traffic load nears the network's throughput limit, retransmissions stop being a solution and start becoming a source of congestion.
Figure 6(b) illustrates that the standard deviation of PSNR increases with higher retry limits for certain congested scenarios (like 2-hop 2500 Kbps), indicating jittery and inconsistent video quality. Furthermore, high retransmission limits in 4-hop scenarios were found to cause end-to-end delays exceeding five seconds, rendering interactive video applications unusable despite any marginal gains in image quality.
Conclusion and Future Directions
Link-layer retransmissions are a double-edged sword in WMNs. While essential for basic reliability under light loads, they can trigger congestion collapse when the network is near capacity. The research suggests that a fixed, high retransmission limit is often counterproductive for multi-hop video streaming. Future work should explore adaptive retransmission strategies that dynamically adjust limits based on real-time congestion levels, or selective retransmission schemes that prioritize critical video data like I-frames to optimize the quality-latency trade-off.
Summary updated