Mesh Capacity WG

1
Mesh Capacity WG
2
Outline

• Technologies
• What band(s) should we use?
• What antennas should we use?
• What architecture should we use?
• Closed platforms are a pressing problem
• Goals What do we need?
• Evaluation
• Repeatability is a major concern
• What should we be repeating?

3
Frequency choice

• Design space is huge
• What radio technology gives what data rate in
  what propagation environment?
• Do we need to lobby FCC for lower frequency
  range, or can we achieve our goals without this?
• Channel and source coding schemes
• Includes rate selection, error correction,
  multipath resilience
• Many permutations are unexplored in a systems
  context
• Additional complications space-time codes, etc.

4
Antennas and Scalability

• Steerable Antennas work outside, not indoors
• Would a simple SVD replacement to steering be
  enough for indoors?
• Transmitter power control yields what gain?
• How much better is MIMO?
• Cognitive radios yield what gain?
• Wideband communication is still poorly understood
  in many ways!

5
Architecture

• Closed platforms are a pressing limitation!
• Everyone wants hardware where they can do
  arbitrary interesting things, but keep everything
  else
• Promising approaches
• Software radios high end that can emulate a
  large variety of low ends
• FPGAs/ASICs (a la Berkeley proposal)
• How do we define interfaces that allow
  cross-layer innovation?

6
What do we need?

• 1Mbps/user would rock!

7
Lack of repeatability is a major concern

• Is simulation necessary for benchmarks?
• Would a new model be enough to achieve fidelity?
  Is anyone making progress on this?
• Does physical emulation yield a better tradeoff
  between repeatability and realism?
• E.g., recent CMU research using DSPs between the
  wires
• Is a testbed necessary?
• If so, does this need to be a shared investment?
  What scale is interesting? What radios are
  interesting? How to achieve repeatability?
• Innovation at all levels of protocol stack ?
  benchmark needs to allow change at all levels
• E.g., try a new MAC/PHY/both
• We need both a simulator that is validated
  against reality and testbeds.

8
What benchmarks are good?

• Strawman proposal
• In-home Mesh
• Community Mesh (neighbor to neighbor)
• 2 Tropos scenarios
• End users funneling into ITAP
• Police cars funneling into ITAP
• PacketHop scenario
• Incident command center
• Darpa scenario
• Tanks and Helicopters and things
• What other scenarios?