Statslab talk

1
Tussle in Wireless Mesh Networking Novelty or
Practicality?
Wenjun Hu
2
(A PhD Students Apology)
3
The Reality

• Wireless mesh networks are promising
• Little requirement on infrastructure
• Ease of configuration and maintenance
• Also seen as a hype
• Fundamental constraint from a shared medium
• Existing protocols unsatisfactory
• So, lets tweak protocols

4
New spin Network coding

• The COPE approach
• A coding shim on the forwarding path
• Let the network be intelligent!
• (Katti et al, XORs in the Air Practical wireless
  network coding, SIGCOMM 2006)
• Observe shim approach is suboptimal
• Joint design of, e.g., coding/MAC, coding/routing
  could be better
• Intelligent network may not fit current layering
  approach

5
Yet more reality

• Need testbed results to validate protocols
• Need existing hardware/software to build testbeds
• Whats available?
• Hardware cards implementing 802.11, typically no
  access to firmware/HAL
• Software standard networking stack (TCP, UDP,
  etc.)

6
Whats the problem?

• MAC artifacts
• Medium allocation not really fair
• Broadcast without collision avoidance
• TCP in wireless
• Backs off over aggressively
• Sensitive to timing, i.e., delay/jitter at the
  medium
• Lots of new protocol proposals in literature, but
  nothing in reality!

7
Solutions?

• The Haggle way
• New architecture from scratch
• Completely rewrite the interfaces/drivers
• Several research groups, many professors,
  researchers, PhD students
• The lone PhD students approach
• Frown/Sigh at existing kit
• Attack a small piece of victim code for
  replacement/extension

8
Tussles, questions

• Brave new idea, or the messy reality?
• Pursue a grand vision of a high performance
  wireless mesh network?
• Or take one stab at a time, get some results
  quickly, and complete one PhD thesis?

9
Answers?