A city-wide vehicular infrastructure for wide-area wireless experimentation

Suman Banerjee, Justin Ormont, Jordan Walker, Suman Banerjee 0001, Ashwin Sridharan, Mukund Seshadri, Sridhar MachirajuWintech 2008 · DOI

TL;DR

This paper introduces a novel city-wide vehicular testbed for wide-area wireless experimentation, leveraging city buses equipped with dual cellular and WiFi interfaces. This infrastructure allows for continuous, low-cost monitoring of wireless network performance, providing valuable insights into user experience and network characteristics under real-world mobility conditions.

What Problem Does This Paper Address?

Monitoring and characterizing the performance of large-scale, wide-area wireless networks, especially those involving mobile users, presents significant challenges. Traditional approaches using static monitoring nodes are inadequate for capturing the dynamic nature of these networks, particularly the impact of mobility on user experience. Furthermore, deploying and maintaining a dense network of monitoring infrastructure is prohibitively expensive. This paper addresses the need for an efficient, scalable, and cost-effective solution to gather fine-grained performance data across a broad geographic area, accounting for diverse mobility patterns.

Methods

The core of the proposed solution is a "client-assisted monitoring" approach. This involves equipping mobile client devices with the capability to collect and report network performance data. Specifically, the authors developed a vehicular testbed consisting of:

  • Dual-Interface Wireless Nodes: Laptops mounted on city buses, each equipped with both a cellular (EV-DO) and a WiFi interface. These nodes also include GPS receivers for location tracking.
  • Software Platform: A suite of software modules running on the client devices that continuously monitor various network parameters. These modules include:
  • GPS Location Extension: Records location data, crucial for associating measurements with specific places.
  • Signal Strength Extension: Monitors cellular signal strength.
  • ICMP Latency Extension: Measures round-trip times for ICMP packets to assess network latency.
  • TCP 1MB Extension: Records throughput for downloading a 1MB file to gauge data transfer speeds.
  • TCP 1 Sec Extension: Captures per-second network interface statistics (bitrates, packet rates) for a more granular view of bandwidth utilization.
  • Mobile Access Router (MAR): To incentivize the city bus system's cooperation, the testbed nodes also provide free WiFi internet access to bus passengers. This MAR system utilizes the EV-DO interface to haul passenger traffic to the internet.
  • Backend Database: A MySQL database serves as a central repository for all collected data. A Perl script parses incoming logs, associates them with GPS coordinates, and stores them for analysis.

This approach leverages the inherent mobility of buses to cover large areas and the ubiquity of client devices to reduce infrastructure costs.

Datasets

The paper introduces a new dataset collected from the city-wide vehicular testbed in Madison, Wisconsin, between March 17 and May 10, 2008. This dataset comprises over 500 hours of mobile and 350 hours of stationary measurements of the 3G EV-DO cellular network. The data includes:

  • GPS location and speed information.
  • Cellular signal strength.
  • ICMP latency.
  • TCP throughput.

This dataset is unique in its scale, mobility coverage, and the combination of wireless technologies monitored.

Key Insights and Ideas

  • Vehicular Testbeds as a Scalable Solution: City buses, with their predictable routes and extensive coverage, offer a cost-effective and practical platform for wide-area wireless experimentation.
  • Client-Assisted Monitoring is Feasible and Beneficial: Shifting the monitoring burden to client devices reduces the cost of infrastructure and naturally captures performance from the user's perspective.
  • Mobility Significantly Impacts Performance: The collected data highlights the variability of signal strength, latency, and throughput across different locations and mobility states, underscoring the importance of mobility-aware network analysis.
  • GPS Data is Crucial but Challenging: While essential for associating measurements with location, obtaining reliable GPS data within enclosed vehicles proved to be a significant challenge, with a high rate of failed localization attempts.
  • Dual-Interface Nodes Offer Richer Insights: Combining cellular and WiFi data allows for a more comprehensive understanding of the wireless environment and potential interferences or performance trade-offs.

Key Results

  • Extensive Data Collection: Over 1.5 months, the testbed collected a significant number of samples for various metrics, including ping tests, TCP downloads, signal strength, and GPS data.
  • Mobility Characterization: The testbed successfully captured vehicular speeds ranging up to 45 mph, with an average bus speed of 14 mph.
  • Signal Strength Mapping: A city-wide RF coverage map was generated, showing variations in signal strength across Madison.
  • Latency and Throughput Maps: Latency measurements typically ranged between 150-400 ms across the city. Average mobile bandwidth was 94 KBps, and stationary bandwidth was 100 KBps.
  • Repeatability: The testbed demonstrated the ability to collect repeatable measurements at stationary locations, providing insights into temporal variations in latency.

Limitations and Future Work

  • GPS Reliability: The authors acknowledge challenges in obtaining consistent GPS fixes within the buses, leading to a significant number of failed localization attempts.
  • Resource Constraints: As the testbed expands to include more resource-constrained devices like cellphones, managing power and bandwidth limitations for data collection will become critical.
  • Data Volume: Scaling the testbed to a larger number of nodes will generate a massive amount of data, requiring efficient sampling strategies and data management techniques.
  • Future Work:
    • Augmenting testbed nodes with Software Defined Radio (SDR) interfaces for agile channel adaptation.
  • Expanding the testbed to a larger scale, potentially incorporating hundreds of buses.
  • Deploying monitoring capabilities on handheld devices (cellphones, PDAs) to capture data from a wider user base.
  • Investigating sampling strategies for spatial dimensions to manage data volume.
  • Incorporating handheld device constraints (power, bandwidth) into data collection mechanisms.
  • Leveraging the MAR system to study user behavior and network traffic patterns.

Summary updated

BibTeX
@inproceedings{banerjee2008city,
  author = {Suman Banerjee and Justin Ormont and Jordan Walker and Suman Banerjee and Ashwin Sridharan and Mukund Seshadri and Sridhar Machiraju},
  title = {{A city-wide vehicular infrastructure for wide-area wireless experimentation}},
  booktitle = {Wintech},
  year = {2008},
  doi = {10.1145/1410077.1410080},
}
Figure