Process Mediation in an Extended Roman Model

1
Process Mediation in an Extended Roman Model

• Dr. Gösta Grahne and Victoria Kiricenko

Concordia University, Montreal, Canada
2
Outline

• Process Mediation
• Existing Approaches
• Our Extended Model
• Roman Model / Process Algebra
• Orchestration and Client Satisfaction
• Algorithm
• Practical View
• Conclusions and Future Work

3
Web Service Process Mediation

• A mediator is a software module that provides
  sharing of services and agglomeration of
  resources into complex services
• Process mediation in Web Services

Web Services
?
Clients
4
Web Service Process Mediation

• A mediator is a software module that provides
  sharing of services and agglomeration of
  resources into complex services
• Process mediation in Web Services

• Discovery
• Composition
• Invocation
• Monitoring

5
Web Service Process Mediation

• We concentrate on
• How to model Web Services
• How to orchestrate / compose

• Verification
• Implementation

6
Existing Approaches

• Modeling and design
• Process Description Languages
• State Machines
• Andrea Ferrara ICSOC04
• LOTOS ? BPEL

• Roman Model
• Daniela Berardi et. al. ICSOC 2003,
• FSM
• Composition reduced to DPL

7
Roman Model
8
Containment
9
Containment
10
Simulation

• A Web Service A can be simulated by a Web Service
  B, denoted A ? B, if there a exist a simulation
  relation R
• Starts states are in R
• For every pair of states (p, q) in R, if p (in A)
  can perform some action and move to some other
  state p, then state q (in B) can also perform
  this action and move to some state q such that
  (p, q) in R
• For every pair of states (p, q) in R, if p (in A)
  is an accepting state then so is state q (in B)

11
Simulation
12
Simulation
13
Web Services Community

• Actions of the individual services can be
  interleaved
• Actions can be executed in parallel
• Execution of a and b perceived as c

14
Web Services Community
15
Web Services Community
16
Web Services Community
17
Web Services Community FSM
18
Satisfaction of Client Requests

• Compute the Web Service Community FSM
• Compute the simulation relation between the
  client FSM and Community FSM
• Problem
• Inefficiency
• Better solution
• Compute the required simulation relation on the
  fly

19
Algorithm

• Based on the partial depth first approach used
  for on-the-flay verification of behavioral
  equivalences
• Jean-Claude Fernandez and Laurent Mounier, CAV
  1991
• Computes the states of the community FSM as
  needed
• the transition relations
• the ? function
• Starts with the pair of start states
• Traverses the FSMS based on the transitions of
  the client FSM
• If a state is encountered second time, the
  algorithm backtracks and constructs the
  simulation relation

20
Algorithm

• Optimistic assumption is used
• If found a counter example to the assumption used
  has to restart
• Runs in EXPTIME
• in the size of transition relations of individual
  services

21
To Sum Up, Our Approach

• Extends Roman Model
• Brings it closer to the reality and allows to
  model more features
• Allows non determinism
• BPEL ltpickgt
• Very common in web service setting
• Allows for parallel execution
• BPEL ltflowgt
• Very important fro distributed environment
• More fine-grained notion of equivalence
  (containment)
• Based on PA
• Uniform Model for Design, Specification,
  Verification, and Automatic Composition
• Formal tools for verification
• Extendable to communions with message passing

22
Future Work

• The further extensions to the model that we are
  currently investigating
• Input / output actions
• Value passing
• Querying / modifying external databases
• We are also investigating the various complexity
  questions related to the composition of the Web
  Services