DataWarp: Building Applications which Make Progress in an Inconsistent World

1
DataWarp Building Applications which Make
Progress in an Inconsistent World

• Peter Henderson,
• Robert John Walters,
• Stephen Crouch,
• Qinglai Ni

2
Background

• Computer systems no longer operate in isolation
• Data is dispersed
• Data volume makes errors inevitable
• Fundamental attitude to data is unchanged

3
Classical view of data
4
DataWarp

• Inspired by TimeWarp
• (Where applications work without a consistent
  notion of the time)
• A framework to allow applications to work without
  fully consistent data

5
DataWarp (2)

• Accept that data errors happen
• Work with what you have
• Make assumptions if you have to
• Be prepared to retract or replace actions

6
DataWarp (3)

• Keep all data
• Construct candidate views of data
• Select a view to use
• Be prepared to change view

7
DataWarp (4)
8
MQDefence

• Experimental Defence Scenario
• Ships on a grid
• Limited ability to discover information about
  environment
• Able to communicate
• Each side seeks to dominate
• the gird

9
MQDefence Playing

• Sink your enemies
• What about unidentified ships?
• Could be allies
• Could be hostile (and dangerous)
• Early decision gives an advantage

10
MQDefence Winning

• Unidentified ships may be hostile or not
• If hostile, then firing immediately gives us an
  advantage
• Should it identify itself as a friend, missile
  could be destroyed in flight
• Shoot first, ask questions later

11
MQDefence Winning

• In DataWarp terms
• A pair of possible views for each unidentified
  ship
• Ships gain an advantage if they use a view in
  which unidentified ships are marked as hostile
• View changes can lead to realisation that an ally
  has been attacked

12
MQDefence - Results
13
Conclusion

• Modern applications are bound to encounter
  shortcomings in their data
• Eliminating these is not a realistic option
• We offer using a DataWarp outlook towards data as
  a solution

14
(No Transcript)