1. Validating Reconfigurations of Reo Circuits
Maurice H. ter Beek1, Fabio Gadducci2 and Francesco Santini3
1ISTI-CNR, Pisa, Italy
2Dipartimenti di Informatica, Università di Pisa, Italy
3EPI Contraintes, INRIA, Rocquencourt, France 2

2. Intro and Motivations Dynamic reconfigurations of Reo circuits
Reo circuits formally model the coordination among components
Graph transformation techniques formally model reconfigurations
Reconfiguration is triggered by events
In our scenario, when system buffers become congested
Applied to a critical scenario in Finance (Credit Suisse)
The e-banking infrastructure can start loosing orders (i.e. money)
Different infrastructures can be considered to save money
A first simulation was presented at ESM’12 (Risk Analysis)
This research was presented at ISARCS’13 (Architecting Critical Systems)

3. Simulation
Beside a formal definition of coordination (Reo) and graph rewriting (DPO), we performed a simulation, to check how reconfiguring works in practice
Modelica (for complex physical systems) and Dymola tool

4. Outline Introduction Reo Graph Transformation
Modelica Simulation Language
Simulation
Conclusion
Future Work

5. In a Nutshell
Reo
Modelica
Graph Rewriting
Simulation

6. Reo [Arbab’04]
A graphical language: it implements the composition of channels into circuits
A circuit describes exogenous coordination
Components (software) are not aware of how interaction works
Circuits made of nodes and (open-ended set of) channels
Nodes are where different channels are synchronized
Components write/take data to/from circuits’ boundary nodes
Every circuit defines an interface through its boundary nodes
Semantics (also) based on constraint automata [Baier et al.’06]

7. Reo channels

8. Example with Writer/ReaderB = qo q1
{A}
fifo = data
{B}
data = fifo s

9. Examples
1) a XOR 2)
2 lossy-fifo

10. Algebraic Graph Transformation
Rewrite (parts of) circuits by manipulating graphical structures
Already proposed for Reo in [Krause’11] and [Krause et al.’13]
Set of rewrite rules that state the objects to be modified
L  R (pattern graph  replacement graph), K invariant/gluing graph
Search occurrence of pattern graph L in G, then replace it to obtain H
DPO approach: a rule is a pair of morphisms l: L  K, r: K  R in the graphs category denoting how graph L is rewritten into graph R while intermediate graphs store connection between items of LHS and RHS

11. Case1

12. Case 2

13. Rewriting Rule 1 K L R

14. Rewriting Rule 2

15. Rewriting Rule 3

16. Modelica
Modelica is an object-oriented, declarative, modeling language for component-oriented modeling of complex systems, e.g. systems containing mechanical, electrical, electronic, hydraulic, thermal, control, electric power or process-oriented subcomponents

17. Modelica (2) An object oriented, modeling language
Modelica classes are not compiled in the usual sense, but they are translated into objects which are then exercised by a simulation engine
The primary content of classes are equations: x + y = 3 – z
They express equalities (not assignments)
The simulation engine discovers the order of execution and which components in the equation are inputs and which are outputs
Algorithmic parts can be present as well
Open (OpenModelica) and commercial front-ends
Dymola: Dynasim AB (now part of Dassault Systèmes)

18. Case 1 (Modelica)

19. Case 2 (Modelica)

20. Aggregating Values

21. Rewriting (Modelica) fifoCongestion > α 1) open 2) case1or2

22. (switch from case1 to case2)
Simulation
fifoCapacity = 100
When
(fifoCongestion ≥ 10%)
then
(switch from case1 to case2)
10K orders, 1h

23. Conclusion and Future Work
Formal design of coordination among components
Easy to verify (e.g. model check) behaviour of coordination
Formal design of event-triggered rewriting of coordination
Simulation through an equation-based simulation language
Application to a critical e-banking system
To stress the infrastructure and have feedback on e.g. risk analysis
Test other coordination schemes
A link to Control Theory and Hybrid Simulation

24. Thank you for your time! Contact/questions: 25