TL;DR
This paper tackles the challenge of implementing network coding (NC), a technique that boosts throughput in wired networks, within the constraints of wireless 802.11 networks. By modifying both the MAC and PHY layers of the 802.11 protocol and implementing these changes on a Software Defined Radio (SDR) platform, the authors demonstrate significant throughput gains (20-30%) in a two-way relay scenario, even with the inherent complexities of wireless communication.
What Problem Does This Paper Address?
Network coding (NC) is a powerful technique that improves network throughput by allowing intermediate nodes to combine packets before forwarding them, rather than just relaying them. While NC has shown great promise in wired networks, its application to wireless networks, particularly the ubiquitous 802.11 standard, presents significant challenges. Wireless networks are inherently different due to their shared medium, half-duplex nature, and susceptibility to interference and channel attenuation. Existing NC research often assumes ideal conditions like scheduled access (TDMA) and symmetric traffic, which don't reflect real-world 802.11 networks that use random access (CSMA/CA) and often experience asymmetric traffic flows. This paper aims to bridge this gap by designing, implementing, and evaluating NC in a practical wireless setting, specifically a two-way relay network using 802.11, and addressing both MAC and PHY layer modifications.
Methods
The authors propose and implement modifications to both the MAC and PHY layers of the 802.11 protocol stack to support network coding.
- MAC Layer Network Coding: They adapt the standard 802.11 MAC layer to support NC by introducing an "NC layer" within the link layer. This layer handles packet queuing and XORing. A key challenge addressed is handling asymmetric traffic and ensuring that the relay node can correctly combine packets from different sources. They introduce the concept of "generation ID" in packet headers to ensure that only packets from the same generation are combined, mitigating potential errors. The implementation uses a three-node relay setup (A-R-B) where nodes A and B communicate via relay R.
- PHY Layer Network Coding: To address asymmetric traffic more effectively, they extend the Decode-and-Forward with Joint Modulation (DF-JM) scheme to the physical layer. This involves modifying the OFDM PHY layer to encode and modulate combined packets at the relay node. They also explore optimal constellation labeling techniques to maximize the gains from side information at the receivers. The implementation uses an OFDM-based PHY layer similar to 802.11a.
The entire system is implemented and tested on a Software Defined Radio (SDR) platform, specifically Microsoft's SORA, which allows for flexible implementation of wireless protocols.
Datasets
Not discussed in the paper. The experiments are conducted using a custom testbed setup with personal computers and SDR hardware.
Key Insights and Ideas
- NC is Viable in CSMA/CA: The paper demonstrates that network coding can provide significant throughput improvements even in the presence of the random access CSMA/CA MAC protocol, which is a departure from prior work often assuming scheduled TDMA.
- Generation ID is Crucial for MAC NC: The introduction of a "generation ID" in packet headers is a practical solution to ensure correct packet combination at the relay node, especially when dealing with the asynchronous nature of wireless transmissions and potential buffer overflows.
- PHY Layer NC for Asymmetry: The DF-JM scheme at the PHY layer proves effective in handling asymmetric traffic, offering substantial throughput gains that are not achievable with MAC layer NC alone in such scenarios.
- Optimal Labeling Matters: The choice of constellation labeling at the PHY layer can significantly impact performance, with optimal labeling providing nearly 100% gain over standard Gray mapping in certain asymmetric scenarios.
- SDR as a Powerful Tool: The use of the SORA SDR platform highlights its utility for rapid prototyping and experimentation of advanced wireless communication techniques.
Key Results
- MAC Layer NC: Achieved a throughput increase of 20-30% compared to a baseline 802.11 network without network coding in a two-way relay scenario. However, this came at the cost of increased packet loss rates, which could be mitigated by larger buffer sizes, albeit at the expense of increased delay.
- PHY Layer NC: For asymmetric traffic, the implemented DF-JM scheme at the PHY layer demonstrated a thirty percent throughput increase compared to a traditional 4-step TDMA exchange. Optimal labeling provided significant gains, especially at lower SNRs for asymmetric relaying.
Limitations and Future Work
- Symmetric Traffic Sensitivity: The paper acknowledges that MAC layer NC is sensitive to traffic symmetry. When one node's power was halved, packet loss increased, and throughput decreased, particularly for the node experiencing reduced power. This highlights the need for PHY layer NC to handle such asymmetries.
- Hidden Node Problem: While the experimental setup aimed to mitigate hidden node issues by placing nodes in an equilateral triangle, the authors note that estimating the impact of hidden nodes is out of scope for this paper.
- Buffer Management Tradeoffs: The paper discusses the tradeoff between buffer size, packet loss, and delay for MAC layer NC. Further optimization of buffer management strategies could be explored.
- Future Work: The authors suggest exploring further optimizations for MAC layer NC in asymmetric scenarios, investigating the impact of hidden nodes, and extending the PHY layer NC implementation to more complex wireless environments. They also mention the potential for further improvements by combining MAC and PHY layer NC more effectively.
Summary updated