This document proposes a method to synthesize a decentralized system from a specification of Live Sequence Charts (LSCs). It introduces distributed LSCs (dLSCs) which use partial orders and local states instead of global states. A two-step process first synthesizes a Petri net structure from LSC main charts, then adds guards to tokens based on LSC precharts. This extracts autonomous components while preserving behaviors. A prototype tool demonstrates the approach on an emergency management example. Future work includes code generation and increasing dLSC expressiveness.

1. Dirk Fahland
Amir Kantor
From LSC
to Distributed Components
Weizmann Institute
of Science

2. Distributed systems: common & complex
web services business processes
wireless
embedded systems
sensor networks
1

3. Design a distributed system
several components
2

4. Create a model for each component
model
describes how the component
works: behavior + interface
model model
model of the system
3

5. More intuitively: describe interactions
X Z a scenario
X  a course of related actions
(across components)
Y Z
specification of the system
4

6. The research problem
synthesize
X Z
specification: system model:
a set of scenarios components
satisfies exhibits
system
behavior
5

7. Current situation for LSC

X Z centralized system
[Bontemps et al. 07]
specification:
a set of LSCs
components
6

8. Live Sequence Charts
Lifeline
event
prechart
message main chart
semantics: if (and when) the prechart occurs,
then the main chart occurs.
7

9. LSC and distributed components
A
B C
trace:
8

10. LSC and distributed components
A
a
B C
trace: a
9

11. LSC and distributed components
A
a
b
B C
trace: a b
10

12. LSC and distributed components
A
a
b
B C
c
trace: a b c
11

13. LSC and distributed components
A
a a
b
B C
c
a
trace: a b c a
a conflicts the LSC
12

14. LSC and distributed components
A
a a “block d”
b
B C
c
a
trace: a b c a
a conflicts the LSC: LSC is no longer relevant
 abort chart  block ‘d’
13

15. LSC and distributed components
A
a a coordinate
b
B C
c
a
trace: a b c a
a conflicts the LSC: LSC is no longer relevant
 may C execute d?
14

16. LSC and distributed components
A
a a coordinate
b
B C
c
 LSC semantics maintains global state
 requires additional interaction or central controller
15

17. Contribution
X Z decentralized system
extract
specification:
a set of LSCs
with
non-global
states
components
16

18. Central Fragment of LSC
 Partial-order representation of LSCs event a by
X Y
component X
X.a
a
causal
b c X.b Y.c
dependency
X Y Y.c
c
d Y.d
a
X.a
17

19. Semantics: distributed LSC (dLSC)
 construct partially ordered runs
if run ends with a pattern
pre
that matches the prechart run
main
then concatenate pre
the main chart locally
main
 preserves causal dependencies in runs
 needs no global state
 no other auxiliary notions
18

20. dLSC Semantics – Example
L1: L2: L3:
Y.c Y.c
X.a
Y.d Z.d
X.b Y.c
X.a Z.e
initial run: X.a
X.b Y.c
concurrent
events
19

21. dLSC Semantics – Example
L1: L2: L3:
Y.c Y.c
X.a
Y.d Z.d
X.b Y.c
X.a Z.e
X.a
X.b Y.c
Y.d
local
X.a concatenation
20

22. dLSC Semantics – Example
L1: L2: L3:
Y.c Y.c
X.a
Y.d Z.d
X.b Y.c
X.a Z.e
X.a X.a
X.b Y.c X.b Y.c
Y.d Z.d
alternative
X.a Z.e
run
21

23. dLSC Semantics – Example
L1: L2: L3:
Y.c Y.c
X.a
Y.d Z.d
X.b Y.c
X.a Z.e
X.a X.a
X.b Y.c X.b Y.c
Y.d Z.d
continue
while
X.a Z.e
possible
X.b Y.c
22

24. Complex precharts
C.
treat
if prechart occurs at the end of the run
 continue with main chart
23

25. Outline
X Z decentralized system
extract
specification:
a set of dLSCs
with
non-global
states
components
24

26. How hard is synthesis?
X Z Petri net
 for each low-level Petri net
specification: exists a dLSC spec with the same runs
a set of dLSCs  cannot (effectively) translate all dLSC specs
to Petri nets, reason: complex precharts
25

27. …need tokens with memory
history
X Z Token History
Petri net
 each token records history of past events
specification:  transitions have guards wrt. token
a set of dLSCs histories
 can express complex precharts of
dLSC
26

28. Synthesis in 2 steps
main charts  net structure
precharts  guards
history
X Z Token History
Petri net
separate nodes
specification:
a set of dLSCs
 polynomial time complexity
components
 complete for local choice specs
27

29. Larger Example: Emergency Management
28

30. Step 1: Synthesize Decentralized System
29

31. Step 2: Group nodes by components
30

32. Result: Extracted components
31

33. Prototype Tool (SAM)
 Components are extracted as a CPN-Tools model
• Supports simulation & analysis
• http://www.win.tue.nl/~dfahland/tools/sam/
32

34. Conclusion
X Z  dLSC = central fragment of Token History PN
LSC with non-global states
 more expressive than extract
low-level Petri nets
dLSC  synthesis + prototype tool
components
33

35. Possible Research Directions
 Code generation (for components)
• Erlang actors
• Java components/threads
 Operations on scenarios
• Scenario (de)composition
• scenario refinement
 Increasing expressive power of dLSC
• mandatory (hot) events
• data
• etc.
34

36. Dirk Fahland
Amir Kantor
From LSC
to Distributed Components
Weizmann Institute
of Science