Introduction
The proliferation of mobile devices has created a digital divide between "connectivity haves"—those with direct access to high-speed gateways—and "connectivity have-nots" who are just out of range. Multi-hop mobile networks offer a promising solution by allowing devices to share their internet links. However, this cooperative model faces two significant challenges: technical limitations inherent to wireless physics and the social problem of node selfishness. This paper presents a decentralized middleware approach designed to enforce fairness and encourage resource sharing in these environments.
The Problem of Selfishness and Interference
In multi-hop wireless chains, the node closest to the internet gateway (the "first node") typically consumes the vast majority of available bandwidth. This "First-Node-Wins" phenomenon is caused by signal interference and packet collisions, which naturally disadvantage more distant nodes.
Beyond technical constraints, social selfishness is a major hurdle. Users are often reluctant to share their battery and bandwidth if it negatively impacts their own device's performance. Existing solutions often rely on a "global view" of the network, which is computationally prohibitive for mobile devices and generates excessive management traffic.
Decentralized Regional Fairness Management
To address these issues, the authors propose an extension to the MMHC (Multi-hop Multi-path Heterogeneous Connectivity) middleware. The core innovation is Regional Fairness Management, which operates on a decentralized principle: nodes only monitor and exchange information with their immediate one-hop neighbors. This keeps overhead extremely low while allowing for effective local control.
Monitoring and Control Mechanism
The system tracks two primary metrics to detect starvation and selfishness:
- Traversing Load (TL): The ratio of forwarded traffic to locally generated traffic.
- Output Effectiveness (OE): The percentage of local packets successfully transmitted.
When a node detects starvation in a neighbor, it issues a "Remote Starvation" (RS) signal. This signal propagates up the chain to the bottleneck node, which then uses standard Linux tools—iptables and tc (Traffic Control) with Hierarchy Token Bucket (HTB) queuing—to throttle its own local applications, thereby reserving bandwidth for the community. The system also rewards "peer connectors" by granting them higher priority weights in bandwidth allocation.
Experimental Evaluation and Results
The researchers evaluated the middleware using ns-2 simulations to model complex multi-hop scenarios and real-world field tests with various portable devices (laptops and smartphones) running Linux, Windows, and MacOSX.
Key Findings
- Throughput Decay: Experiments quantified the rapid drop in performance over distance. While a 1-hop path achieved 5.09 Mbps (802.11b), throughput plummeted to 0.73 Mbps by the 7th hop due to cumulative interference and collisions.
- Fairness Imbalance: Without intervention, a 2-hop multi-client test showed extreme disparity, with one node receiving 4.30 Mbps while the other received nearly zero.
- Effective Throttling: The middleware successfully equalized consumption. By implementing dynamic
tcrules, the system could shift bandwidth allocation (e.g., from an 80/20 split to 50/50) to ensure all nodes maintained functional connectivity even under saturated conditions.
Conclusion and Future Work
The proposed middleware demonstrates that regional monitoring can effectively manage fairness in multi-hop networks without the overhead of global coordination. Future research will focus on trust management to prevent malicious nodes from spoofing starvation signals and improving session continuity for streaming media during network reconfigurations.
Summary updated