Stephan Reiff

1
Hybrid Techniques for Detecting and Resolving
Feature Interactions in Telecommunication Services
Resolution Choices for an Online Feature Manager
FIW00 Glasgow, 16 -19 May 2000

• Stephan Reiff
• University of Glasgow
• http//www.dcs.gla.ac.uk/sreiff
• sreiff_at_dcs.gla.ac.uk

2
Work, Project and Partners

• Joint work with Prof. Muffy Calder
• Project (EPSRC funded)
• Hybrid Techniques for Detecting and Resolving
  Feature Interactions in Telecommunication
  Services
• Partners
• University of Strathclyde, Glasgow
• Citel
• Mitel
• http//www.dcs.gla.ac.uk/research/hfig/

3
Overview

• Motivation and Background
• The Hybrid Approach
• Definitions
• Feature Interference
• The Feature Manager
• Solution Space
• Open Questions and Further Work

4
Motivation and Background

• Motivation
• Feature Interaction Problem
• Legacy and Proprietary Systems
• Extension needed
• Knowledge about internal behaviour
  unavailable/unreliable
• Background
• Work by Marples,Magill and Tsang all have
  shortcomings
• We use Marples and Magills approach
• Hybrid?
• Use offline analysis to find general rules ...
• to control online feature manager

5
Example Features (1)
6
Definitions

• Messages
• IO-aspect(event, source, destination, value)
• relations between messages
• equal, inverse and name equivalent
• Alphabet
• Set of messages (partitioned into Input and
  Output messages)
• Trace
• non-empty, finite sequence of messages (starting
  with an input message)
• Feature
• non empty set of traces over a minimal alphabet
• Show examples

7
Example Features (2)

• iaf (forwardcall,id,n,a)-(oinform,id,a,"Call
  Forwarded")
• cfu (ialert, a, b,-)-(forwardcall,id,n,a)
• tl (offhook, id, id,-)-(announce,id,id,"PIN?
  ")
• (onhook, id, id,-),
• (offhook,id,id,-)-(tau),
• (offhook,id,id,-)-(announce,id,id,"PIN?")
• (dial,id,id,n)-(tau),
• (offhook,id,id,-)-(announce,id,id,"PIN?")
• (dial,id,id,n)-(announce,id,id, "wrong PIN")
• (onhook,id,id,-)

8
Feature Interference
Definition Feature Interference n features
F1Fn with alphabets a1an interfere iff In
words Features interfere iff they have either a
common input message or one features output
message is the other features input message.
9
Resolution Process
10
Solution Space (1)

• What is it?
• A set of all traces that can be generated from
  the features individual traces
• simple example
• How do we obtain it?
• Add all traces of just one feature
• Add all combinations of traces of 2 features
• Add all combinations of traces of 3

11
Solution Space (2)

• The combinations are sets of all possible
  overlapping interleavings
• Here we will consider overlapping interleavings
  of just 2 traces
• example

12
Examples for oi2

• Example 1
• t1 (a-bc) t2 (x-a-b)
• oi2(t1, t2) (x-aa-bc)
• Example 2
• t3 (d-e) t4 (d-f)
• oi2(t1, t2) (d-e-f), (d-f-e)
• Question Why keep -aa?
• Answer Examples (CFB BC with busy and BC
  CND with free)

13
Pruning Simple Rules

• No useful behaviour is provided by
• consecutive announcement messages without an
  intermediate input
• tones and announcements after onhook
• The message t can be removed from traces
• Duplicate traces can be removed
• Example

14
An Example

• gtgtgt printshort(fm("xxx",fiwtl,fiwtcs))
• These features interfere,finding resolution ...
  
• 0 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,tau )
• 1 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,wrong PIN )(-,tau )(,onhook ,-)
• 2 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,wrong PIN )(,onhook ,-)(-,tau )
• 3 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,tau )(-,announce,wrong PIN )(,onhook ,-)
• 4 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,screened )(,onhook ,-)(-,tau )
• 5 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,screened )(-,tau )(,onhook ,-)
• 6 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,tau )(-,announce,screened )(,onhook ,-)
• 7 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,wrong PIN )(-,announce,screened )
• (,onhook ,-)(,onhook ,-)
• 8 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,wrong PIN )(,onhook ,-)
• (-,announce,screened )(,onhook ,-)
• 9 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,screened )(-,announce,wrong PIN )
• (,onhook ,-)(,onhook ,-)
• 10 (,offhook ,-)(-,announce,PIN? )(,dial,n
  )(-,announce,screened )(,onhook ,-)
• (-,announce,wrong PIN )(,onhook ,-)
• 11 (,offhook ,-)(-,announce,PIN? )(,onhook
  ,-)
• 12 (,offhook ,-)(-,tau )

15
Open Questions and Next Steps

• We need to refine pruning
• What are bad resolutions?
• What is a/the best resolution?
• Is there always a best resolution?
• Can we combine construction and pruning?
• What is the complexity?
• How well does the approach scale up?
• Is an observable approach sufficient?
• Use Process Algebra?
• New Operator(s) for Construct and Prune

16
Conclusion
17

• Any Questions?
• References
• http//www.dcs.gla.ac.uk/research/hfig
• http//www.dcs.gla.ac.uk/sreiff

18
Relations between Messages

• Name equivalent, event type and values are equal
• Equal , name equivalent and IO aspect are equal
• Inverses, name equivalent and IO aspect not equal

19
A functional View of the FM

• The feature manager has to construct the solution
  space and extract the best solution from it.
• Definition Feature Manager
• Explain individual subfunctions

20
Construct
21
oi2
First(tj) denotes the first message in a trace.
22
Overlapping Interleaving
23
f() and int()