Native XML Business Process Execution for Computer Supported Mobile Adaptive Business Processes

1
Native XML Business Process Executionfor
Computer Supported Mobile Adaptive Business
Processes

• Thomas Hildebrandt
• associate professor, ITU
• Komialt seminar, ITU, Dec 5th, 2006
• FTP research project 2007 - 2010
• Kjeld Schmidt, Henning Niss and Mikkel
  Bundgaard (ITU)
• Microsoft Development Center Copenhagen

2
Road Map

• The Computer Supported Mobile Adaptive Business
  Processes (CosmoBiz) research project
• Traditional Business Process Models and execution
• Challenges for Pervasive Business Processes
• Native XML BP execution and formalisation

3
Computer Supported Mobile Adaptive Business
ProcessesITU Microsoft Development Center
Copenhagen

• Interdisciplinary research project combining
  industrial prototyping and research in formal
  models and computer supported cooperative work
• Starts january 1st, 2007 and runs for 4 years.
• 2 PhDs and 1 year postdoc funded by the danish
  research agency for technology and production
• Relates to 2 industrial PhD projects Danske Bank
  (MDBP) and Resultmaker (Clinical Workflows)

4
Work packages
5
Models of Business Processes
Model-based design, e.g. from graphical
flow-chart notation to XML-based execution
language
BPMN
WS-BPEL
6
Traditional architecture
The WfMC reference model
7
Business Process Execution

• business processes are long-lived
• process state is persisted
• - traditionally in a proprietary relational
  format
• The informal semantics is hidden in the process
  execution engine

8
Challenges for Pervasive BPM
How do we support

• context-dependent, mobile business process
  instances ?
• (static) guarantees for correctness and security
  ?
• (higher-order) processes for process management
  ?
• evolution and adaption of business process
  languages ?

9
Our working thesis

• Business process instances must be exchangeable,
• the execution semantics must be formalised to
  support consistent execution on different devices
  and advanced programming language concepts such
  as types, mobile and higher-order processes,
• and the formalisation must be extensible and
  close to the language and its implementation to
  support language evolution, engineering (e.g.
  exploring new features based on CSCW field
  studies and formal models) and execution in
  practice

10
A possible solution

• Formalise process execution using theory of
  graph-rewriting and process calculi, implemented
  as XML-rewriting and persisted in XML-store
• Proof of concept A formalized native XML
  execution of BPEL implemented on top of a
  peer-to-peer XML persistence layer

Described in two ITU MSc thesis projects and
presented at the International workshop for
models and tools for coordination (MTCoord) in
2005 and the COORDINATION conference in 2006
11
Graph rewriting systems

• A graph rewriting system consists of a definition
  of the valid graphs and a set of rules of the
  form (L?R) specifying that a sub graph L can be
  replaced with the sub graph R
• Example Petri Net can be seen as an instance of
  a graph rewriting system
• Christian Stahl has provideda complete Petri
  netsemantics for BPEL 1.1

graphical extensive tool support, but not
easily extensible, nor as close to the language
as one could hope for!
12
Process Calculi

• A process calculus is a formal textual notation
  for (usually concurrent communicating) processes
  equipped with rewrite rules for execution and a
  theory of relating processes (e.g. simulation)
• ExampleThe pi-calculus is a famous example of a
  process calculus for communicating processes with
  dynamic communication channels
• Lapadula, Pugliese and Tiezzi have proposed a
  WS-process calculus - supports WSDL type
  system, but not easily extensible

13
Motivation and Challenges

• Business process execution languages evolve
  quickly
• Specifications are ambiguous and implementations
  are released fast...
• Can we provide an extensible formalization that
• support model-based development, implementations
  and formal verification?
• can be extended and adapted when the languages
  evolve?
• will support distributed, mobile and adaptive
  business processes - and also in practice?

14
Existing formalisations

• Christian Stahl has provided a complete Petri net
  semantics for BPEL 1.1 graphical, extensive
  tool support,but not easily extensible,nor as
  close to the language as one could hope for!
• Lapadula, Pugliese and Tiezzi have proposed a
  WS-process calculus - supports WSDL type
  system, but not easily extensible
• Reiko Heckel is working on graph-rewriting
  formalisations

(from slide of C. Stahl)
15
Bigraphical Reactive Systemsand Reactive XML

• A bigraph is a collection of trees (the place
  graph) with links (the link graph)
• Bigraphical Reactive Systems (Robin Milner,
  2001)is a graph rewriting framework (and
  theoryof simulation) for bigraphs inspired by
  thepi-calculus but also able to describe Petri
  Net
• Incidentally, bigraphs correspond closely to XML
  data and thus the syntax of BPEL, e.g.

16
Reactive XML as Bigraphical Reactive Systems

• The XML processes correspond to pure bigraphs
  (Milner, CONCUR 2001), e.g.
• A bigraph is a collection of trees (the place
  graph) with links (the link graph)
• A BRS is specified by a signature (node labels)
  and a set of reactions
• - a graph-rewriting framework inspired by
  pi-calculus and Mobile Ambients, equipped with a
  general theory of contexts and composition, and
• labelled bisimulation congruence (Milner Hoeg
  Jensen, POPL03) (not yet explored)

instancenametransfer. (variables.
(variablenameaccountA.P1
variablenameaccountB.P2 P3)
sequence. (assign.copy.(fromvaraccountA

tovaraccountB) P)
denotes
17
Our approach

• Exploit close correspondence between Bigraphical
  Reactive Systems (BRS) and XML to get
• an extensible formalisation as a BPEL calculus
• implementation as Reactive XML
• distributionmobility (peer-to-peer XML Store)
• Business process language, formalisation and
  implementation closely connected - and extensible
• Spin-off from Bigraphical Programming Languages
  (BPL) project (funded by ITU and the Danish
  Research Agency)

18
Native XML BPEL execution as (bi)graph rewriting

• The idea Combine the process description and
  state and execute by rewriting as in theory of
  process calculi and graph-rewriting

persist, distribute and exchange process
instances as XML
19
- a word on the XML persistence layer (XMLstore)

• never destroys data, but shares subtrees, e.g
• maintains complete history (e.g. for
  transactions) and easy replication (e.g. for P2P
  distribution)

20
XML process calculus

• we use a shorter process calculus notation (maps
  1-1 to XML)

maps to
21
The BPEL subset

• A system consists of a set of processes and
  instances in parallel
• Instances are defined as processes except
  allowing values for variables

22
Systems, Processes Instances

• A system consists of a set of processes and
  instances in parallel
• Instances are processes allowing values for
  variables

23
Formalising computation steps

• computation steps can be described as bigraphical
  reactions
• specified by a set of parametric rewrite rules
  L?R (e.g. parametric in the variable names,
  content and context)

24
Parametric Reactions Contexts
25
Examples of reactions

• Semantics of conditional
• Semantics of while

26
Implementation Reactive XML
27
Motivating Higher-order Contexts

• assign action and variables should belong to same
  instance

28
Higher-order Contexts
29
Rewrite rule examples

• Semantics of assignment (in process calculus
  notation)
• Semantics of exit

30
Context constraints

• use XPath to constrain the set of contexts that
  can be placed in a higher-order hole

?
31
Semantics II

• Semantics of process invocation (invoke/receive)
• again using higher-order contexts to ensure
  correct scope of variables
• (correlation considered in MSc thesis of M.
  Olsen)

32
From BPEL to Bigraphs back
33
Summary

• XML rewriting framework called Reactive XML
  formalised as Bigraphical Reactive Systems and
  distributed P2P implementation using value-based
  XML persistence layer (MTCoord Workshop, 2005)
• BPEL case process language, formalisation and
  implementation closely connected - and
  extensible(Coordination Conference 2006)
• ITU Technical Report TR-2006-85.(extended,
  submitted journal version available)

34
The architecture for pervasive BPM?
35
PerspectivesExperimentalPlatform for reliable
distributed (business) process execution
coordination
36
Perspectives Future Work

• Extensible platform for distributed (business)
  process language engineering, execution, CSCW
  coordination
• Future (partly covered by the CosmoBiz project)
• Full BPEL semantics and implementation
• Model-based development (i.e. transformations
  from BPMN/UML to BPEL)
• Typed, mobile and higher-order processes
• CSCW studies in adaptive, context-dependent,
  mobile business and work flow processes, e.g.
  sales services and clinical workflows
• Formal reasoning verification
• Quantitative semantics Time and probabilities
  (with M. Kwiatkowska)
• Relate to other graph rewriting frameworks (R.
  Heckel, graph rewriting)

37

• Thank you for staying to the end- now lets
  have the lottery -)

38
Extensible Process Calculus
39
Motivating higher-order contexts

• Recall how instances bind values to variables
• Reaction rule needs to pick the correct instance
  variables
• How do we deal with the context parameter C ?

40
Higher-order contexts

• Higher-order contexts allow us to abstract the
  context C in
• and write
• expecting a parameter

41
Higher-order interfaces

• Technically, the hole of the context in the
  parameter
• should be reflected in the outerface, and the
  nested process in the redex
• should be reflected in the innerface

42
Higher-order interfaces

• We restrict our self to higher-order contexts of
  type
• given by the grammar

43
XPath constraints

• Add an Xpath constraint to higher-order holes
• Used to restrict to contexts for which the
  location of holes are specified by the XPath
  constraint, e.g. active contexts are specified
  by

44
Perspectives and open ends...

• Reliable Extensible Implementation of Distributed
  (peer-to-peer) Business Process Execution
• Full category of higher-order contexts
• Higher-order and Mobility?
• Formal Reasoning/Verification?
• Context dependency?
• Types?

45
Business Process Execution

• long-lived persistent processes bank loan,
  sales services, production workflows, ...
• supported by workflow or business process
  management systems(as part of an ERP system)
• use XML based process execution languages
  (WS-BPEL) and graphical flow-chart notations
  (BPMN)

Workflow Management System Reference Model
46
The mobility challenges

• Client and business process execution engine
  mobility
• Business process mobility
• embedded in mobile client
• mobile subprocesses
• Issues
• Disconnected operation, disconnected
  coordination...
• Peer-to-peer and ad-hoc networking
• Location, role, ..... context-dependency
• Higher-order processes as values

Break the boundaries...!
47
Mobile Salesmen

• Client mobility Salesmen visit customers
• Process mobility - a sales service process may
  be delegated- the check record subprocess may be
  moved to the sales process (to allow disconnected
  operation)
• Context-dependency Sales service depends on the
  current customer and salesman
• Higher-order A sales process may be modified to
  reflect new or personal practices
• .......BREAK THE BOUNDARIES.....

48
Coordination and CSCW challenges

• How is mobile, distributed work and business
  processes coordinated ... and carried out by
  humans?
• Make use of roles, location, situation, context...

49
Reliability challenges

• Process execution languages evolve quickly
• Current specifications are ambiguous and
  implementations are being released fast...
• Messages are typed (with WSDL and XML Schema) but
  what about type checking...?
• How can we make reliable implementations that can
  be extended to support reliable mobile and
  higher-order processes?

50
Background Theory

• Formal models and types for higher-order mobile
  embedded processes, the Homer calculus
  (Hildebrandt, Bundgaard)
• Computer Supported Distributed and Mobile Work
  (Schmidt)
• Type systems and value-based peer-to-peer XML
  storage (Niss)
• Extensible formal process models and operational
  semantics (bigraphs) (Hildebrandt, Niss,
  Bundgaard, Robin Milner)

51
Background Implementation

• Microsoft Business Solutions Mobile (Röser,
  Gufler)
• Research-based implementations of reliable
  it-systems (Niss et al)
• Peer-to-peer BPEL with Bigraphs and Reactive XML
  (Hildebrandt, Niss, Olsen - COORDINATION 2006,
  Bologna, June, 2006)
• ABACO and Ariadne Towards a Technology of
  Coordination (Schmidt et al)

52
Selected Related Work

• The 3gERP, NEXT and BEFORE research projects
• Process Engines Bizztalk, ActiveBPEL,
  Websphere...
• BPEL-SPE (Extension for sub-processes) (white
  paper IBM and SAP, 2005
• Models Petri Net, CCS and pi-calculus for
  Business and Workflow Processes (Wil van der
  Aalst, Chr. Stefansen, A. Farrell,...)
• Formalised coordination frameworks and workflow
  engines, eg. - the COORDINATION conference
  series - Andrew Farrell, Imperial, - Wil van
  der Aalst Technical Universität Eindhoven
• Typed graph-rewriting (König, Montanari, ...)
• Types for WS-BPEL (COORDINATION 2006)

53
Student project thesis ideas

• Implement peer-to-peer BPEL execution engine
  based on Reactive XML (see Business Process
  Execution with Bigraphs and Reactive XML, ITU
  Technical report TR 2006-85)
• Mobile BPEL (client/engine or subprocesses)
  implementation and/or CSCW field studies, e.g.
  BPEL for the Mobile Salesman
• Context-dependent BPEL
• Types for BPEL