COS-598A, Spring 2017, Kyle Jamieson

Lecture 3: Roofnet

Context, ca. 2000–2005

Today: Roofnet

Roofnet: Design Choices

  1. Volunteer users host nodes at home

  2. Omnidirectional rather than directional antennas

  3. Multi-hop routing, not single-hop hot spots

Roofnet: Goals and non-goals

Roofnet deployment

Node addressing

Internet gateways

Links: Wired v. wireless

Example: Varying link loss rates

Hop count and throughput (1)

Minimum-hop-count routes are significantly throughput-suboptimal

Hop count and throughput

Link loss is high and asymmetric

Routing protocol: Srcr (1)

Routing Protocol: Srcr (2)

Link metric: Strawmen

ETX: Expected Transmission Count

Measuring link delivery ratios

Multi-bitrate radios

ETT: Expected Transmission Time

ETT: Assumptions

  1. Underestimates throughput for long (≥ 4-hop) paths
  2. Overestimates throughput when transmissions on different hops collide and are lost

Auto bit-rate selection


Roofnet evaluation

Wide spread of end-to-end throughput

End-to-end throughput by hop count

Comparing with computed throughput

Forwarding indeed creates interference

User experience: Mean throughput from gateway

What link ranges/speeds to expect?

Which network links does Srcr use?

Lossy Links are Useful

Diversity in node use: “Meshness”

Why not Access Points?

Evaluation strategy: Multi-hop v. AP

Optimal AP (GW) placement

Random AP (GW) placement

Roofnet: Concluding thoughts