Hybrid Techniques for Detecting and Resolving Feature Interactions in Telecommunications Services Mi

1
Hybrid Techniques forDetecting and
Resolving Feature Interactions in
Telecommunications ServicesMid-Project
ReportJanuary 2000Muffy Calder and Stephan
Reiff Department ofComputing ScienceUniversity
of Glasgow
2
Overview

• Project Goal
• Glasgow Approach
• Role of Reasoning
• Results
• Questions
• Future Work

3
Project Goal
terminal
terminal
terminal
Off-line analysis
Legacy switch
4
The Glasgow Approach
Legacy switch
Components
Feature manager
Interaction resolution generation and
selection
Legacy switch
Components
Feature manager
Model is prescriptive, not just descriptive.
5
The Initial Model
User
User
User
User
...
switch
...
feature
feature
Basic call
Basic call
feature
feature
Basic call
Basic call
feature
feature
feature
feature
6
A Next Model (?)
User
User
User
User
...
switch
Feature manager
Feature manager
Feature manager
Feature manager
...
feature
feature
Basic call
Basic call
feature
feature
Basic call
Basic call
feature
feature
feature
feature
7
Interaction Resolution

• Assume detection is done.
• FM receives more than one response.
• What do we know about the system?
• Only what we can observe.
• What are possible resolutions?
• Generate them. Non-trivial.
• Eliminate bad resolutions.
• What is bad?
• Define best resolutions.
• Greatest or just maximal?
• Select one.

f1
f2
Simple case
f1
f2
f2
f1
8
Role of Reasoning

• high level - validation of model, feature
  manager approach
• generic properties deadlock, livelock
• specific, temporal properties
• eg. computation path, user i.
• ((dial,i) -gt (busy_tone))
• low level - feature manager behaviour
• specification of (re)solution space
• calculation of (re)solution space
• complexity of algorithms

9
Results
Legacy switch
Components
Feature manager

• Modelling Legacy Telecommunications Switching
  Systems for Interaction Analysis (MC, SR)
• SEBPC book
• Towards Strong Resolution Mechanisms for an
  Online Feature Manager (SR)
• FIW00 submission
• Model Checking Networks of Telecomms Services
  using SPIN (MC, AM)
• report

Interaction resolution generation and
selection
Legacy switch
Components
Feature manager
10
Some Questions

• Should we have multiple instances of feature
  managers? Do we need a meta-level feature
  manager? What is the right architecture?
• Can a theory of events be rich enough to prune
  solutions space? Are there any generic
  properties? How should pruning be expressed,
  regular expressions/automata c-f grammars
  worse?
• Can high level reasoning scale up? Under what
  circumstances does P(n) gt P(n1)? Is n
  dependent upon number of features or number of
  users, or both? How can we define underlying
  recursively defined set?
• Can generation and pruning be dovetailed
  effectively?
• Will features with unbounded looping be
  required? E.g. do we really need automata (as
  opposed to traces)?

11
Future Work

• Specify pruning, selection, and associated
  algorithms.
• Induction.
• Iterate.
• Apply to DESK or live system.