Counting Interface Automata and their Application in Static Analysis of Actor Networks

1
Counting Interface Automata and their
Application in Static Analysis of Actor Networks
7. September 2005 SEFM, Koblenz, Germany
E. Wandeler(1), J.W. Janneck(2), E.A. Lee(2), L.
Thiele(1)
(1)Computer Engineering and Networks
Laboratory ETH Zurich, Switzerland (2)Department
of EECS UC Berkeley, USA
2
Introduction (1) Actors

• An actor is a SW/HWcomponent with
• Input Ports
• Output Ports
• Parameters
• State
• When an actor is fired, it
• consumes data-tokens on its input ports
• produces data-tokens on its output ports
• updates its state

AddAndAcc
1
0
7
3
5
3
7
3
Introduction (2) Actor Models
Director ? - Control flow
What is the semantic of this model?
Receiver ? - Communication
Actor A
Actor C
Actor B
4
Motivation forStatic Analysis of Actor Models

• Actor-oriented modeling is widely spread, e.g. in
  embedded system design.
• Compatibility-checks of the components in an
  actor-model at design-time reduce modeling
  errors.
• Static analysis leads to formal verifiable
  results.

5
Analysis-Strategy
Composition ? - Component compatibility
Analysis ? - Legal firing-sequence - Buffer
overflow - Deadlock - ...
Actor Automaton
MoC Automaton
1.
2.
3.
Composition
4.
Analysis
6
Counting Interface Automata
7
Counting Interface Automata
States (Normal, Initial, Reset)
8
Counting Interface Automata
Transitions
9
Counting Interface Automata
step f! fR?
Actions (Input ?, Output !, Internal )
10
Counting Interface Automata
Ports (Input, Output)
11
Counting Interface Automata
Counter Variables (Persistent/Transient)
12
Counting Interface Automata
Tan Tbn
cTn
Counter Declarations/Assignments
13
Counting Interface Automata
Tgtn / hTT!
Guards
14
Counting Interface Automata
hT!4
Action Quantities
15
Counting Interface Automata
hT!4
hT?n
Action Quantity Declarations
16
Composition

• Valid Compositions
• a! a?
• a!3 a?3
• a! a?1
• a!3 a?n -gt n3
• Invalid Compositions
• a! b?
• a!3 a?2

17
Example
actor A () In gt Out A1action a gt a
end end
actor Src () gt Out1, Out2 A1action gt
a, end A2action gt , a end
selector (A1A2) end end
actor Snk () In1, In2 gt A1action a,
b gt end end
actor B () In gt Out A1action a gt a,
a end end
18
An Actor Model Example
19
Step 1 Actor Automaton
1.
20
An Actor Model Example
actor B () In gt Out A1action a gt a,
a end end
21
Step 1 Actor Automaton
actor B () In gt Out A1action a gt a,
a end end
22
Step 2 MoC Automaton
1.
2.
23
Step 2 MoC Automaton
IncTgtn / Ing?n
24
Step 3 Composition
1.
2.
3.
25
Step 3 Composition
MoC
Actor
26
Incompatible Components
MoC
Actor
IncTgt1 false!
27
Incompatible Components
MoC
Actor
empty
28
Model Automaton
1.
2.
3.
29
Model Automaton

• Empty, if incompatible components exist in the
  actor model
• Contains otherwise
• all actor-connections
• all token-exchange-rates
• state-space of the model
• firing-schedule, if existing

30
Step 4 Analysis
1.
2.
3.
4.
31
Token Exchange Automaton
32
Token Exchange Petri Net
33
Petri Net Analysis

• Transition-Invariant IT0
• ? No legal firing sequence
• Reachability tree
• SnkIn2T is unsecure, i.e. buffer-Overflow in
  input In2 of Snk
• no deadlock

34
Conclusions

• CIA are taylored towards the use in actor
  systems.
• Methods exist to extract CIA from CAL actor
  definitions.
• Methods exist to generate CIA for SDF DDF MoC.
• Successful composition if CIA behavioral type
  compatibility!
• CIA of full model contains much information for
  further static analysis.

35
Thank you!
36
Example
actor A () In gt Out A1action a gt a
end end
actor Src () gt Out1, Out2 A1action gt
a, end A2action gt , a end
selector (A1A2) end end
actor Snk () In1, In2 gt A1action a,
b gt end end
actor B () In gt Out A1action a gt a,
a end end
37
Inkompatible Komponenten
MoC
Actor
IncTgt1 ist false!
38
Token Exchange Automaton
39
Token Exchange Petri Net
40
Petri Net Analysis

• Transition-Invariant IT0
• ? No legal firing sequence
• Reachability tree
• SnkIn2T is unsecure, i.e. buffer-Overflow in
  input In2 of Snk
• no deadlock

41
Erreichtes

• Definition von Counting Interface Automata (CIA)
• Automatische generation von Actor CIA aus der Cal
  Actor Language (CAL)
• Automatische generation von MoC CIA für SDF und
  DDF
• Automatische komposition von kompletten Actor
  Modellen
• Beispiele von Analyse-Möglichkeiten

42
Zukünftige Arbeit

• Automaten für weitere Models of Computation
  definieren
• Erweiterte Analysen für den Modell Automat
  entwickeln
• Benutzung der vorgestellten Techniken zur
  Code-Generierung
• Benutzung von Counting Interface Automata für
  andere Anwendugsgebiete

43
An Actor Model Example