Dynamism in BPEL

1
Dynamism in BPEL

• Onyeka Ezenwoye
• Latin American Grid project.
• Monday, August 28, 2006

2
The need for Dynamism

• Fault Tolerance
• Network failure, resource availability, etc
• Optimization
• Responsiveness to real-time events, resource
  scheduling, etc
• Variations in user input
• Multiple configurations for experiments.

3
Approaches to dynamism(1)

• Fool the System
• Actions are performed outside the control of the
  workflow.
• Workflow unaware of changes.
• Usually through some middleware
• Message redirection, dynamic binding
• May use high-level policy
• E.g. TRAP/BPEL 1

4
Approaches to dynamism(2)

• Modify running workflow instance
• Task may be cancelled, delegated or redirected.
• Other instances of the same workflow are not
  affected.
• Usually requires some dependency list and/or
  policy specification
• E.g. ScyFlow 2, TRIANA 3

5
Approaches to dynamism(3)

• Modify workflow definition
• New behaviour specified in Workflow definitions.
• May migrate running instances to new ones.
• Migration requires careful management
• Difficult to encapsulate every possible
  orchestration response.
• Administrative overhead.

6
BPEL and dynamism

• BPEL is NOT dynamic!
• Adequate semantics (pick, skip tasks)
• Hierarchical but not modular.
• May implement process as several interacting
  workflows (cumbersome).
• Minimal fault handling.
• Supports compensation handling.
• Dynamism through middleware
• application, messaging.

7
References

1. Onyeka Ezenwoye and S. Masoud Sadjadi. Trap/bpel
   A framework for dynamic adaptation of composite
   services. Technical Report FIU-SCIS-2006-06-02.
2. Karen M. McCann et. al. ScyFlow An Environment
   for the Visual Specification and Execution of
   Scientific Workflows.
3. Matthew Shields and Ian Taylor Programming
   Scientific and Distributed Workflow with Triana
   Service.
4. Niels Joncheere et. al Requirements for an
   Aspect Oriented Workflow system for Grid
   Services.
5. Fabio Casati A Discussion on Approaches to
   Handling Exceptions in Workflows.
6. Rashid Hamadi et. al. Recovery Nets Towards
   Self-Adaptive Workflow Systems.