OWL-P: Processes = Protocols + Policies Munindar P. Singh ( - - PowerPoint PPT Presentation

OWL-P: Processes = Protocols + Policies

Munindar P. Singh (Students: Amit K. Chopra, Nirmit V. Desai, Ashok U. Mallya)

singh@ncsu.edu

Department of Computer Science North Carolina State University http://www.csc.ncsu.edu/faculty/mpsingh/

c Singh et al., November 2004 – p.1/29

Problem, Solution, Approach

Problem: Modeling and enacting open business processes Traditional approaches can’t handle autonomy, heterogeneity, dynamism Incorporating context is essential Solution: Interaction is the key New way of thinking geared toward open systems Approach: Protocols capture interaction Software engineering: refine, aggregate protocols Agents: flexible enactment Compliance in the face of flexibility

c Singh et al., November 2004 – p.2/29

Accomplishments: Intellectual

Protocols provide interaction-centric modeling, leaving policies to participants Commitment semantics yield flexible modeling and enactment Theory of protocols supports reusability, refinement, and aggregation of interactions Students Three PhD dissertations being supported PhD dissertation involving autonomic service selection based on OWL for QoS being defended on Dec 6.

c Singh et al., November 2004 – p.3/29

Accomplishments: Implementational

On SemWebCentral OWL-P as an OWL ontology (using SWRL) Roles Messages: content as propositions and commitments Rules to describe messages and roles Protégé plugin for OWL-P protocol editor Protocol composer and skeleton generator Agent-based architecture layered on FIPA Rule-based policies that help agents satisfy their protocol roles

c Singh et al., November 2004 – p.4/29

Accomplishments: Evangelical

Papers, tutorials, panels, invited talks Trying to reach the software engineering community: well-received at OOPSLA Contact with IBM and HP Beginning project jointly with IBM on autonomic and agent based business process management New book: Service-Oriented Computing: Semantics, Processes, Agents IEEE Internet Computing track (2005) on Service-Oriented Computing

c Singh et al., November 2004 – p.5/29

The Essential Tension

Reusability requires Context freedom Encapsulation Usability (usefulness) requires Context sensitivity Varieties of context include organizations, laws, and the real world Main idea Autonomy: components have a life of their own Interactions are what matter

c Singh et al., November 2004 – p.6/29

A Process is . . .

Orchestration: a partial order of actions under the control of a central conductor Akin to a workflow or flow in BPEL Choreography: an exchange of messages among participants Akin to a conversation as described by WS-Chor Collaboration: a joint set of activities among business partners Akin to real business; essential for SOAs

c Singh et al., November 2004 – p.7/29

Emphases of Collaboration

Commitment Protocols: Content & Compliance Rule-Based Commitment Protocols: Flexibility Protocols: Modularity I m p l e m e n t a t i

• n

a n d e n a c t m e n t M

• d

e l i n g a n d v a l i d a t i

• n

M

• n

i t

• r

i n g a n d c

• m

p l i a n c e Dynamic Organizations

c Singh et al., November 2004 – p.8/29

Innovations: 1

Protocols: Conceptually decentralized, reusable, encapsulations of processes Commitments: Content for protocols Support reuse via abstractions for refinement and aggregation of protocols What the protocol should accomplish What deviations are legitimate and what aren’t Operational semantics for commitments

c Singh et al., November 2004 – p.9/29

Innovations: 2

Rule-Based Reasoning: Expressing protocols flexibly Accommodating context Deciding specific actions by applying policies Spheres of Commitment: Modeling organizations Enacting protocols Monitoring and verifying compliance Slogan: Processes = Protocols + Policies

c Singh et al., November 2004 – p.10/29

Trends and Assessment

Increasing # of business protocols IOTP , Escrow, SET, NetBill, . . . RosettaNet: 107 Partner Interface Processes (PIPs) ebXML Business Process Specification Schema (BPSS) Intended to be legally binding Generally highly limited: two party, request-response protocols No commitments; no formal semantics Limited support for modeling or enactment

c Singh et al., November 2004 – p.11/29

Vision

Target Audience: Practitioners Formalization in the background Engineering: not full automation, but tools for Modeling and validation of protocols Modeling and validation of processes Generation of software components Enactment via Spheres of Commitment Monitoring and compliance

c Singh et al., November 2004 – p.12/29

Usage Scenario

Order OWL-P Shipping OWL-P Payment OWL-P Axioms

Software Designer Composer

Purchase OWL-P Local Policy Protocol Repository

specify register MERCHANT

Merchant Skeleton OWL-P Merchant Local Process

+

UDDI Repository

C U S T O M E R Search Merchant Merchant Port

1 2 3 4 5

lookup

6 7

Purchase.Customer Skeleton

10 8 9 Customer Skeleton OWL-P Local Policy + Customer Local Process

register

11

c Singh et al., November 2004 – p.13/29

Simple Scenario and Example Run

A customer (C) looks up a book at a vendor (B) and is quoted price and availability C orders the book from B B ships to C C pays B

reqQuote(c,b,g) sendQuote(b,c,g,p) sendAccept(c,b,p) sendMoney(c,b,p)

s0 s1 s2 s3 s4 s5

Bookstore, b Customer, c sendGoods(b,c,g)

c Singh et al., November 2004 – p.14/29

Process View: Flow or Protocol

Customer

Select Send Receipt Ship Pay

Bookstore Shipper Bank Customer Bookstore Customer Bookstore Customer Bookstore Bookstore Customer Bank Shipper

Ship Pay Pay Select Order Ship

c Singh et al., November 2004 – p.15/29

Challenges: Modeling

Refinement: pay by credit card versus pay Extensibility: verify C’s attributes, e.g., age Adjustment: receive payment before shipping; receive book before paying Alternative execution examples: B arranges for a shipper (S) to deliver the book to C C pays via bank (K) Compose a process from the above

c Singh et al., November 2004 – p.16/29

Refinement of Protocols

Selection criteria for protocols Functional: pay versus ship Nonfunctional: payer trusts payee or not

Pay Pay with receipt Pay cash Pay via credit card Pay with cash and receipt Pay with check Pay via debit card

c Singh et al., November 2004 – p.17/29

Aggregation of Protocols

A simplified protocol may be revealed to a give role Decisions could be taken internally but not exposed

Bookstore Customer Bank Shipper

Ship Pay Pay Select Order Ship

c Singh et al., November 2004 – p.18/29

Example Run: Pay via Bank

reqQuote(c,b,g) sendQuote(b,c,g,p) sendAccept(c,b,p) sendGoods(b,c,g) authPay(c,b,p) s0 s1 s2 s3 s4 s21 Bookstore, b Customer, c s5 Customer's Bank, k sendMoney(k,b,p)

c Singh et al., November 2004 – p.19/29

Example Run: Shipper Protocol

s10

reqQuote(m,s,[gv]) sendQuote(s,m,[gv],q) sendAccept(m,s,[gv],q)

s11 s12 s13 s13

sendGoods(m,g,s)

s14

Shipper, s Sender, m

s15

sendMoney(m,s,q)

s16

sendGoods(s,v,g) Receiver, v

s15

c Singh et al., November 2004 – p.20/29

Example Run: Composed Purchase

reqQuote(c,b,g) sendQuote(b,c,g,p) sendAccept(c,b,g,p)

s0 s1 s2

s3

s4 s5

authPay(x,p)

s21

sendMoney(k,x,p) reqQuote(b,x,[gc]) sendQuote(x,b,[gc], px) sendAccept(b,x,[gc],px)

s11 s12 s13 s13

sendGoods(b,g,x)

s14

sendMoney(b,x,px)

s16

sendGoods(x,c,g)

Shipper, x Bookstore, b Customer, c Bank, k Shipping Payment

c Singh et al., November 2004 – p.21/29

Challenges: Enactment

Behaving adaptively: decide dynamically to ship before payment to trusted Cs Handling exceptions External problems: cannot ship book Context-sensitivity: not legal for kids Detecting violations: no payment; book arrives damaged Correcting violations: remind, complain, refund, . . . Exploiting opportunities: combine orders from same C

c Singh et al., November 2004 – p.22/29

Example Run: Return and Refund

Example: Uniform Commercial Code (UCC) allows returns with refunds for goods that are received damaged

reqQuote(c,b,g) sendQuote(b,c,g,p) acceptQuote(c,b,p) sendMoney(c,b,p)

s0 s2 s3 s4 s5

Bookstore, b Customer, c

s5 s18

returnGoods(c,b,g) sendGoods(b,c,g) sendRefund(b,c,p)

s19 s1

c Singh et al., November 2004 – p.23/29

Processes = Protocols + Policies

Operational patterns Time outs, remind, garbage collect, . . . Decisions to manipulate: delegate, assign, . . . Enact protocols dynamically based on agent policies and context Transactional patterns Induce transactional scopes Apply retry, redo, undo (compensate) where appropriate Enact via Spheres of Commitment

c Singh et al., November 2004 – p.24/29

Remaining Work: Easy

Challenge Remediation Simplify protocol design for business Libraries

• f

compos- able protocols Produce compliant agents Refined methodology based on policies Make up to date with Rules work and OWL-S Freshen the work

c Singh et al., November 2004 – p.25/29

Remaining Work: Middlish

Challenge Remediation Formalize context Develop operational se- mantics in π-calculus Protocol compliance Apply commitment se- mantics Organizational, trans- actional exception mod- eling Exploit Spheres

• f

Commitment Optimize role selection based on QoS Incorporate service quality representations

c Singh et al., November 2004 – p.26/29

Papers on this Topic

Recent papers in ICWS, AAMAS, OOPSLA, ICSOC address parts of the above vision Tutorials at WWW, AAMAS, OOPSLA Panels at WWW, AAMAS, ICWS “Agent Communication Languages: Rethinking the Principles.” IEEE Computer, 31(12):40–47, Dec 1998 “Reasoning About Commitments in the Event Calculus: An Approach for Specifying and Executing Protocols.” Annals Math & AI, 42(1-3), 2004 “Verifying Compliance with Commitment Protocols.” J. Autonomous Agents & MAS, 2(3):217–236, Sep 1999

c Singh et al., November 2004 – p.27/29

Spheres of Commitment

Buyer Seller Inventory Packaging Shipping Billing Operations

c Singh et al., November 2004 – p.28/29

Contexts as Transformers

g f r f g d d d

r r t

r

g f f g d t d t

c Singh et al., November 2004 – p.29/29