Preemptive Adaptation To Trouble Spots In Wireless Networks

1
Pre-emptive Adaptation To Trouble Spots In
Wireless Networks

• James Pascoe
• Vaidy Sunderam
• Roger Loader

The University of Reading, UK and Emory
University, Atlanta, GA. Work In Progress
session Mobicom 2001. http//www.collaborative-co
mputing.com/
2
Background and Motivations

• Background group communications, notably, in
  Collaborative settings.
• Motivation Rethink group communication protocols
  in the context of mobile networks. Propose model
  and architecture for wireless aware group
  protocols.
• Target platforms Ricochet and 802.11.

3
The Problems

• So, given that Ricochet and 802.11 give us IP
  (albeit wirelessly), what is the issue ?
• Two problems
• Garbled communication tentatively solved by
  reliable wireless transport technologies.
• Intermittent connectivity larger problem
  impacting on vulnerable parts of traditional
  group communications systems.

4
Trouble Spots

• Areas of intermittent connectivity / impaired
  quality.
• Group protocols can not differentiate between
  failed hosts and hosts in trouble spots.
• Group membership algorithms invoked erroneously
  causing eviction of live group members.
• Traditional group communication semantics
  possibly invalidated at application level.

5
Preliminary Study Ricochet

• Initial Thesis
• we can exploit location based information to
  pre-emptively adapt to trouble spots at the
  protocol level.
• Possible resolution strategies include
• Dynamically upgrading Quality of Service
• Increased buffering
• Graceful shutdown
• Warn the user

6
Preliminary Results Response Time
7
Preliminary Results Packet Loss
8
Preliminary Results Signal Strength
9
Preliminary Study Consolidation
10
Caveats

• GPS was used to determine location.
• Network dependant solutions exist.
• Trouble spots are dynamic suggesting a real-time
  monitoring approach.
• Collecting reliable data was difficult.
• Client / server architecture must be adopted
  carefully when connectivity is in question.

11
Next Study

• Purpose built tools RadioTool / RRNAPI.
• Assumption that location information is not
  available. The requirement is for the mobile
  device to know it is entering a trouble spot.
• Metrics re-adjusted measure of scatter in
  response time looks suitable. Packet loss and
  signal quality offer additional confirmation and
  insight.

12
Next Study Results 802.11
13
Next Study Results 802.11