Performance evaluation of a DTN as a city-wide infrastructure network

Joohyun Lee, Kyunghan Lee, Jaesung Jung, Song ChongCFI 2009 · DOI

TL;DR

This paper explores the potential of Delay Tolerant Networks (DTNs) as a city-wide infrastructure network, leveraging vehicles as data carriers. They demonstrate that DTNs can offer superior capacity compared to traditional mesh networks, especially in urban environments with high mobility, by analyzing real-world taxi trace data.

What Problem Does This Paper Address?

Providing robust and high-capacity data networks in urban areas is a significant challenge. Traditional cellular systems can be expensive to deploy and may struggle with peak demand. While wireless mesh networks offer a solution by using multi-hop routers, they still face limitations in terms of installation costs and capacity, especially when dealing with dense traffic. This paper investigates an alternative: using Delay Tolerant Networks (DTNs), where vehicles act as mobile data carriers, to create an efficient and cost-effective city-wide infrastructure network. The core problem is to evaluate the performance and capacity of such a DTN system, particularly in comparison to existing mesh network architectures.

Methods

The authors propose an architecture for a DTN as an inter-hotspot network. In this setup, vehicles move between "hotspots" (popular locations like squares, stations, or department stores) and act as data couriers. They download data from one hotspot and upload it to another when they connect.

Key methodological aspects include:

  • Sectorized Antennas: To overcome the bottleneck of single-point transmissions at base stations within hotspots, the paper assumes the use of sector antennas. These allow for multiple simultaneous transmissions within a hotspot, significantly boosting capacity. The capacity gain is shown to increase with the number of sectors, converging towards an "ultimate capacity."
  • DTN Transport Methodology: An ideal data distribution algorithm is assumed, which has perfect knowledge of node mobility to upload data to vehicles that can deliver it most rapidly. This is a single-copy scheme, meaning no data redundancy is considered.
  • Simulation Framework: The authors developed a DTN simulator in MATLAB using real-world mobility traces. They compare the DTN's performance against a mesh network under similar conditions.

Datasets

The primary dataset used is the Shanghai taxi trace data [14]. This dataset comprises GPS traces from 4063 taxis over 28 days, providing rich information about urban mobility patterns. The authors selected 3234 taxis with complete trace data for their simulations.

Key Insights and Ideas

  • DTNs as a Viable Infrastructure: The paper strongly suggests that DTNs are not just for niche applications (like interplanetary communication) but can serve as a practical and high-capacity infrastructure for urban data networking.
  • Capacity Advantage: The DTN, especially when augmented with sector antennas, consistently outperforms mesh networks in terms of capacity, particularly during peak hours.
  • Time-Varying and Asymmetric Capacity: The DTN's capacity is inherently tied to the ebb and flow of urban traffic. It exhibits significant time-varying characteristics (e.g., higher capacity during rush hours, lower at dawn) and asymmetric upload/download rates, mirroring real-world mobility patterns.
  • Cost-Effectiveness: A significant advantage of the DTN approach is its potential for lower infrastructure costs, as it leverages existing vehicles rather than requiring the deployment of numerous fixed routers.

Key Results

  • Single-Link Capacity: In a single-link scenario between two hotspots, the DTN with 4-sector antennas achieved 4.79 times more capacity than a mesh network during morning rush hour.
  • Multi-Link Capacity: In a multi-link scenario involving four hotspots, the DTN demonstrated an even greater advantage, achieving 5.93 times more capacity than a mesh network under the same conditions.
  • Sector Antenna Impact: The DTN capacity increases significantly with the number of sectors, reaching about 80-90% of the ultimate capacity with 12 sectors.
  • Time-of-Day Performance: DTN capacity is notably higher during morning and evening rush hours and significantly lower at dawn.

Limitations and Future Work

  • Ideal Data Distribution: The current simulation assumes an ideal data distribution algorithm with perfect knowledge of mobility. This is a significant simplification, and developing practical, efficient routing and data dissemination algorithms for DTNs remains a challenge.
  • Delay and Delivery Ratio: While capacity is the focus, the paper acknowledges that delay and delivery ratio are critical issues in DTN research. Future work will analyze these aspects and their relationship with capacity.
  • Feasible Algorithms: The authors plan to develop and integrate feasible data distribution algorithms into their future work.

Summary updated

BibTeX
@inproceedings{lee2009performance,
  author = {Joohyun Lee and Kyunghan Lee and Jaesung Jung and Song Chong},
  title = {{Performance evaluation of a DTN as a city-wide infrastructure network}},
  booktitle = {CFI},
  year = {2009},
  doi = {10.1145/1555697.1555717},
}
Figure