Nonlocal Choice and Beyond: Intricacies of MSC Choice Nodes

1
Non-local Choice and BeyondIntricacies of MSC
Choice Nodes

• Presenter To-ju Huang
• Supervisor Ingolf Krüger

Arjan J. Mooij, Nicolae Goga, Judi M.T.
Romijn, Non-local Choice and Beyond Intricacies
of MSC Choice Nodes. FASE 2005
2
Outline

• Introduction
• Message Sequence Charts
• From MSC to Implementation
• Message Sequence Graphs
• Problematic Choice Nodes
• Dealing w/ Non-local Choice Nodes
• Summary

3
Message Sequence Charts

• Capture system-wide behaviors
• Specify scenarios/requirements easily

4
From MSC to Implementation

• Incompatibility
• MSC captures system-wise behaviors
• Real implementation is process-based
• Naïve implementation is problematic
• Ex. Deadlock, extra behaviors
• Core problem choice nodes

5
Message Sequence Graphs

• hMSC hierarchy w/ depth 1
• For simplicity, choices nodes are made explicit

6
Outline

• From MSC to Implementation
• Problematic Choice Nodes
• Non-local Choice
• Non-deterministic Choice
• Race Choice
• Dealing w/ Non-local Choice Nodes
• Summary

7
Non-local Choice

• Non-local choice leads to deadlock!

choice
msc_base
msc_NLC
8
Non-local Choice (contd)

• Deadlock scenario

9
Non-deterministic Choice

• While message X is received, which MSC to execute?

choice
msc_base
msc_NLC
10
Race Choice

• Z may arrive later than X and Q will make wrong
  decision

choice
msc_base
msc_NLC
11
Outline

• From MSC to Implementation
• Problematic Choice Nodes
• Dealing w/ Non-local Choice Nodes
• Traditional Approaches
• Relaxing
• Solution for 2-process Case
• Solution for N-process Case
• Summary

12
Traditional Approaches

• Localize all choices
• Fail for autonomous processes
• Specify behaviors exhaustively
• Complicated, non-achievable
• Treat in implementation
• With coordination protocols
• Can humans be coordinated?

13
Relax it a Little Bit!

• Allowing extra behaviors
• Keeps what you want, allowing the others if not
  harmful
• Makes systems implementable
• Delayed choice
• Parallelism allowed before decision made
• Converge later to local choice
• Its sometimes domain-specific!

14
2-process Solution

• Constraint branching bMSC must be
• RC-like (Request-Confirm scenario) or
• A-like (Announce scenario)

Arjan J. Mooij, Nicolae Goga, Dealing with
non-local choice in IEEE 1073.2s standard for
remote control,SAM 2004 SDL And MSC, LNCS 3319
15
2-process Solution (contd)

• Constraint choice node must be

16
2-process Solution (contd)

• Choices becomes local for arbiter Q

17
N-process Solution (contd)

• Running Example ATM

Customer cancels before banking system replies
Banking system replies before customer cancels
18
N-process Solution

• Choose an arbiter process which is
• Non-human, early participated, common
• Split each branching MSC into 2 parts
• Front run in parallel with other fronts
• Tail where real choices are made
• After splitting, arbiter is able to make local
  decision in these tails

19
(No Transcript)
20
N-process Solution (contd)

• Constraints for splitting
• Added parallelism is used in next choice

21
N-process Solution (contd)

• Constraints for splitting

no e, if s1 ? s2
if event e
no e
no e
22
Summary

• Its non-trivial from MSC to implementation for
  choice nodes
• With some relaxing,
• Easier to deal with non-local choices
• While main behaviors still hold
• However, relaxing
• Sometimes domain-specific
• Should be checked carefully

23

• Thank you!