Skip to content

Lab 2 — Why a static mesh costs more control traffic than OLSR

Goal. Measure routing overhead as ants per delivered packet. Explain why a protocol built for change keeps paying for it on a network that does not change. Switch off the mechanism that backs off, and measure what it saves. Time. 45 min. Site. The mesh world, seed 8, and the tuning bench.

← Classroom pack · Answer key

Background (5 min)

Normalized routing load (NRL) counts control transmissions per delivered data packet. On the ns-3 static mesh, OLSR delivers 99.82 % at NRL 1.78. AntHocNet delivers 99.34 % at NRL 4.51 (static-mesh.md). AntHocNet keeps sampling routes that never change. Quiet mode (enableQuietMode, #571) backs proactive ants off while a destination's best route stays the same. It is on by default since v2.3.0. Before that, every route was probed every 10 s whether or not anything changed.

1. Predict (5 min)

  1. In the browser mesh, about how many ant transmissions per delivered packet do you expect after 120 s: under 1, about 5, or over 20?
  2. With quiet mode switched off, will that number rise? Will delivery change?

2. Run (15 min)

  1. Open #mesh?seed=8: rooftop routers that never move.
  2. Run to t = 120 s and pause. Record ✓ and ants, then compute ants ÷ delivered. This is the default: quiet mode on.
  3. In the tuning bench, set enableQuietMode to off. Run to t = 120 s again and record the same numbers.

3. Explain (10 min)

  1. By what fraction does quiet mode cut ants per packet (off → on)? Did delivery change?
  2. Quiet mode backs off only while a route stays the same. What would happen on a network where routes keep changing?
  3. Quiet mode still sends some proactive ants. Why not stop them all?

4. Compare with ns-3 (10 min)

The v2.3.0 campaign measured quiet mode on every network family, paired per seed against the v2.2.0 defaults (v2.3.0-campaign.md), and its result moved the default on. Find the static-mesh row, and say whether quiet mode closed the gap to OLSR's NRL.