Nirmit Desai, Munindar P' Singh

1
July 6th, 2004
Protocol-Based Business Process Modeling and
Enactment

• Nirmit Desai, Munindar P. Singh
• Department of Computer Science
• North Carolina State University

2
Outline

• Motivation
• The big picture
• Conceptual model
• Scenarios and examples
• Technical framework
• Skeleton generation
• Compatibility verification
• Prototype implementation
• Future directions

3
Motivation
Limitations of service composition efforts

• Lack of modularity Absence of aggregation /
  inheritance hierarchies
• Inadequate abstractions Monolithic flow models
• Restricted autonomy of parties due to centralized
  modeling
• Difficult to verify and predict the behavior in
  face of exceptions / opportunities

Solving these problems has significant value.
4
Contributions

• Modular approach for capturing business
  interactions as protocols
• Process (public)Protocol(s) (private)Policies
• Modularization enables reuse of protocols as
  components
• Framework for enacting processes
• Generating BPEL flows from published protocol
  specifications
• Verifying compatibility of constituent protocols
• Proof of correctness

5
Contributions (Contd.)

• Separation of concern along the protocol
  boundaries
• Constituent protocols are almost independent of
  one another
• Publishable protocol specifications independent
  of an individual partys local business policies

6
Example Netbill Protocol
1
(C, M) Request
2
(M, C) Offer
3
(C, M) Accept
4
(M, C) Goods
5
(C, M) Payment
6
(M, C) Receipt
7
7
Business Protocols

• Capture the business interaction from the global
  view
• Modeled as finite state machine representation
  (aka P-FSM) of the captured interaction
• States have no explicit semantics
• Serve as publishable, modular components for
  enacting processes
• Useful abstractions facilitating aggregation and
  reuse

8
Terminology

• Protocol Skeleton (aka P-Skel) Protocol as
  viewed by a participant (local view)
• Agent A real-world enterprise / party
• Composite Skeleton (aka C-Skel) One or more
  P-Skels combined together
• Local Flow An instance of a C-Skel augmented
  with local business policies

9
Conceptual Model
Business Protocol
10
Example Supply Chain Process
Supply Chain Process
11
SHIPPER
14. status req
15. status resp
17. delivery
1
1
.

t
4
r
.

a
y
c
3
e
.
k

s
f

e
a
s
t
i
i

b
p
l
i
C
e
h
M

1. ship options
s
?

.
Motivating Example
U
0
E
1

2. ship info
S
R
T
C
5. pay options
O
H
6. pay info
M
A
9
.

r
e
c
e
i
p
t
E
N

1
2
.

s
h
i
p

c
o
n
f
i
r
m

R
1
T
3
.

c
7. auth req
a
p
t
d
u
r
e
e
r
u
K
t
p
O
a

c
h

t
.
u
6
1
A

.
8
PAYMENT GATEWAY
12
Example Escrow Protocol
1
B Buyer S Seller E Escrow
(B, E) deposit
2
(E, S) secured
3
(S, B) goods
4
(B, E) goodsOK
5
(E, S) pay
6
13
Skeleton Generation
Given a P-FSM, how to generate local flows ?

• Partition the P-FSM at role boundaries
• Compute the state of the protocol as viewed by
  each role at each step of the protocol
• Resultant skeletons
• Represent protocol from the perspective of a
  given role
• When executed together, realize the original
  protocol execution
• Transformation is proven sound and complete

14
Skeleton of Buyer
15
Skeleton of Buyer (Contd.)
16
Skeleton of Seller
17
Incompatible Skeletons

• A set of role skeletons generated from a P-FSM
  are always compatible with each other. What if

18
Compatibility Verification

• Central system of versioning is unrealistic for
  open environments
• Run Ping-Pong algorithm before agents start
  interacting
• P-Skels of communicating parties are matched
  against each other
• Each message sent by a party must be received by
  the other party and vice versa
• And, their view of the current protocol state
  must be consistent

19
Prototype Implementation

• Machinery for realizing the conceptual model
• Implementation of skeleton generation and
  compatibility verification
• Grounded in the SOC infrastructure
• Generating local flows as BPEL processes
• Exposing the agent interface via WSDL
• Simplifications
• Splicing of P-Skels into a C-Skel not addressed
• P-FSM is a simplistic representation of a
  protocol

20
Prototype Scenario
Netbill P-FSM
21
Consumer BPEL (Generated)
22
Consumer WSDL (Generated)
23
Relevant Literature

• OWL-S BPEL Modern approaches for service
  composition
• Lack modularity and hence reusability
• Interaction patterns intertwined with local
  policies
• Semantical capabilities of OWL-S are
  complementary
• WSMF A composite service modeling framework
• Process model is similar to OWL-S
• Ontologies and discovery mechanism are
  complementary
• WSDL with WSCI Interface description with
  choreography
• Centralized view for modeling
• Lacks modularity and hence reusability

24
Future Directions

• Semantics of actions
• Actions of the participants have meanings
  attached to them and can be reasoned about
• Mechanically combining more than one skeletons
• Dependencies among skeletons need to be
  standardized
• Richer formalisms to specify protocols
• State-machines are primitive

25
Questions ?
Thanks !!