Overlay Wireless Sensor Networks for Application-Adaptive Scheduling in WLAN

Sonia Waharte, Jin Xiao, Raouf BoutabaHSNMC 2004 · DOI

TL;DR

This paper proposes a novel two-tier architecture for Wireless Local Area Networks (WLANs) that uses a sensor network overlay to manage traffic scheduling. By offloading control plane functions to the sensor network, they achieve contention-free data transmission, leading to improved bandwidth utilization and reduced delay, all while adapting to application-specific Quality of Service (QoS) needs.

What Problem Does This Paper Address?

Wireless Local Area Networks (WLANs), particularly those based on the IEEE 802.11 standard, face a fundamental challenge in providing guaranteed Quality of Service (QoS) for diverse applications. While the standard offers some mechanisms like Distributed Coordination Function (DCF) for fair access and Point Coordination Function (PCF) for centralized scheduling, both have limitations. DCF, while simple, can lead to substantial bandwidth under-utilization and potential starvation of low-priority traffic. PCF, though offering better QoS, is complex to implement and can also be inefficient. The core problem is how to achieve both high bandwidth utilization and guaranteed QoS for applications like multimedia streaming, which are increasingly prevalent in WLAN environments, without introducing excessive complexity.

Methods

The authors propose a two-tier architecture that augments the existing IEEE 802.11 data plane (operating at 54Mb/s) with an overlay sensor network acting as a control plane (operating at 2Mb/s). This separation allows for parallel communication and offloads the overhead of control messages from the data plane.

The sensor network is composed of three types of nodes:

  • Manager: Attached to the Access Point (AP), it acts as a data aggregation server, makes scheduling decisions, and sends replies.
  • Relays: Fixed sensors strategically placed to extend the manager's reach and route messages.
  • Agents: Sensors attached to mobile nodes, responsible for monitoring buffer occupancy and reporting it to the manager.

The key innovation is the Application-Adaptive Scheduling Algorithm. At the start of each transmission cycle, agent sensors report their mobile node's buffer occupancy. The manager then uses this information to schedule the next transmission cycle, allocating time slots based on application needs. This scheduling information is sent back to the mobile nodes before the current cycle ends. The algorithm aims to be fair by ensuring each stream gets its required bandwidth, but it's "application-adaptive" because excess capacity is re-distributed to nodes that need more bandwidth, rather than being wasted.

Datasets

Not discussed in the paper. The evaluation is based on simulation studies.

Key Insights and Ideas

  • Separation of Concerns: Offloading control plane functions (scheduling, monitoring) to a dedicated sensor network overlay is a powerful way to de-congest the data plane and enable contention-free operation.
  • Application-Adaptive Fairness: The concept of "fairness" is redefined to be dynamic and responsive to application demands. Instead of a rigid equal share, bandwidth is allocated based on need, with excess capacity intelligently redistributed.
  • Leveraging Sensor Networks: Sensor networks, with their distributed monitoring and communication capabilities, are well-suited to act as a control infrastructure for existing wireless networks like WLANs.
  • Contention-Free Data Plane: By moving scheduling to the control plane, the data plane can operate without the delays and inefficiencies caused by contention for the medium.

Key Results

  • Reduced Transmission Delay: The proposed architecture significantly outperforms the simple DCF scheme in terms of overall transmission delay. For example, the maximum wait time between packet transmissions is kept under 100ms.
  • Improved Bandwidth Utilization: The architecture achieves higher bandwidth utilization compared to DCF, attributed to the reduction in contention overhead.
  • Scalability: The system demonstrates graceful scalability with the number of nodes, with total registration time remaining under 40ms even for 50+ nodes.
  • Fairness Preservation: The architecture guarantees fairness among users while ensuring application-adaptive QoS.

Limitations and Future Work

  • Complexity of Overlay: While the data plane is simplified, the introduction of a sensor network overlay adds its own layer of complexity in terms of deployment and management.
  • Control Plane Throughput: The control plane operates at a lower throughput (2Mbps), which could become a bottleneck if the number of control messages or the frequency of updates becomes very high.
  • Future Work: The authors suggest extending the architecture to support service class differentiation, cellular-WLAN QoS interworking, and per-traffic QoS guarantees (minimum bandwidth and maximum delay). They also propose exploring better communication techniques for the sensor overlay to improve control exchange efficiency.

Summary updated

BibTeX
@inproceedings{waharte2004overlay,
  author = {Sonia Waharte and Jin Xiao and Raouf Boutaba},
  title = {{Overlay Wireless Sensor Networks for Application-Adaptive Scheduling in WLAN}},
  booktitle = {HSNMC},
  year = {2004},
  doi = {10.1007/978-3-540-25969-5\_61},
}
Figure