Semantically-aided Data-aware Service Workflow Composition

1
Semantically-aided Data-aware Service Workflow
Composition

• Ondrej Habala, Marek Paralic,
• Viera Rozinajová, Peter Bartalos
• Marek.Paralic_at_tuke.sk
• Technical University of Košice, Slovakia

2
Content

• Introduction - Semantics and Web services
• Motivation for the SEMCO-WS
• Architecture of the system for semantic
  composition of services
• Ontologies for Automatic Workflow Management
• Support of Collaboration in the SEMCO-WS
  Environment
• Design of the process of automated data reuse -
  Workflow Composition and Execution Module
• Conclusion and Future Work

3
Introduction

• Semantic Web Services (SWS) extension of Web
  Service technologies with semantics in order to
  automate tasks common when using WS
• Tasks publishing, discovery, selection,
  composition, invocation, deployment and ontology
  management
• SWS frameworks combine semantic description of WS
  capabilities, input, outputs and behavior with
  the syntactic description in WSDL and XML-Schema

4
Semantics and Web Services

• Functional Semantics - supposing that the
  functionality of a WS is fully described by its
  inputs and outputs, we can apply matching
  algorithms
• Data Semantics can overcome restrictions of the
  syntactical and structural description about data
• QoS Semantics - after discovering according to
  the requirements potentially applicable services,
  the selection of most suitable service is needed
• Execution Semantics - involves e.g. flow of
  actions, communication patterns, message
  sequence, preconditions and effects of WS
  invocations
• Domain Semantics - healthcare, bio-informatics,
  telecommunications, military, school
  administration

5
Motivation

• Meteorological and hydrological experiments
• Lot of data is reused in several experiments
• Semantic Composition of Web and Grid Services
  (SEMCO-WS) tries to deal also with the what
  of automated workflow composition
• New environmental risk management application
• Knowledge Based Workflow System for Grid
  Applications (K-Wf Grid)
• Successful, but not everything it was hoped to
  achieve
• Simplified workflow construction
• Interesting problem

6
Architecture of SEMCO-WS
7
Ontologies for Automatic Workflow Management

• Aims for using an ontology
• better data organization by adding additional
  metadata describing semantics of the data
  available in the system (for describing IOPE),
• to conceptualize the domain and provide a
  vocabulary of terms occurring in it (user
  communication and collaboration),
• ontology captures the domain of risk management
  and related areas (concepts from geography,
  meteorology, natural disasters and risk
  management),
• user model that contains information about the
  user's knowledge and relation to the domain
  entities.

8
Ontologies for Automatic Workflow Management (2)
9
Support of Collaboration

• our aim is to provide knowledge-based
  collaboration tool
• first we designed the communication tool
  facilitating the users to communicate in standard
  ways
• then we augmented the tool utilizing the semantic
  information incorporated in ontology - the main
  aspects
• processing on-line discussion
• dividing the discussion according to given
  criteria into the units
• getting the areas of interest that were discussed
• according to these areas of interest offering the
  previous discussions to the given topic or
  providing other relevant documents

10
Workflow Composition and Execution Module

• Petri net model for the workflow description,
• Abstract Workflow Composition Tool (AWCT) is
  responsible for workflow construction and uses
  backtracking from the final activity to the
  initial activities of the workflow,
• Executable Workflow Composition Tool (EWCT) -
  Extended process of workflow refinement that
  changes an abstract workflow into the executable
  one,
• Workflow Execution Engine (WfEE) is the place
  where workflow is running in, it works on
  Petri-net principle of passing tokens,

11
Data Components
URLs of services Descriptions of jobs
Component database
Token database
Token metadata
All existing tokens (both created by users and
produced by jobs/services

• Semantics of each component
• What type of data it produces
• Its input-output transformation description
• Its output-input transformation description

Semantic description of all tokens in token
database based on required data parameters, it
allows to identify the proper token

• For each component
• Transformation of input tokens into input data
• Transformation of output data into output tokens

12
Workflow Construction Process
Component database
Token database
Token metadata
?
?
X
lttoken typemm5outgt ltxgt120lt/xgt ltygt250lt/ygt ltstart
gt2008-07-21lt/startgt lt/endgt2008-07-22lt/endgt ltwidthgt
300lt/widthgt ltheightgt150lt/heightgt lt/tokengt
We have found a service which is able to produce
the required token however, it requires two
input tokens, and the search continues
Now we have found services which need only
existing input tokens so we have a workflow and
input data, and the application may start
We have produced the required output token (which
represents the output data), the workflow has
finished
We start with the semantic description of the
required output
We look for a token which would satisfy the
requirements
No such token is found we look for a service or
job which can produce token of such type
13
Workflow Execution Process
ltinputStructgt ltmapFilegt map317.tif
lt/mapFilegt ltdataFilegt dispData.bin
lt/dataFilegt ltdensitygt 12 lt/densitygt
ltstartTimegt 1280982 lt/startTimegt lt/input
Structgt
ltoutputStructgt lttimeDatagt simul-out.txt
lt/timeDatagt ltanimationgt spread.avi
lt/animationgt lt/outputStructgt
lttimedatagt lttimeDatagt simul-out.txt
lt/timeDatagt lt/timedatagt
ltmapgt ltmapFilegt map317.tif
lt/mapFilegt lt/mapgt
ltdatagt ltdataFilegt dispData.bin
lt/dataFilegt lt/datagt
ltanimgt ltanimationgt spread.avi
lt/animationgt lt/animgt
ltconfiggt ltdensitygt 12 lt/densitygt
ltstartTimegt 1280982 lt/startTimegt lt/config
gt
1. We need input data
2. Input tokens are combined into input data for
the service or grid job
3. This input data is used to call the service,
or execute the grid job
4. The service/job returns output data
5. The output data is divided into tokens these
are annotated by metadata and stored into the
database
14
Prototype implementation of EWCT and WfEE

• The functionality of this module was already
  tested at the text-mining workflow that links
  together text mining services,

15
Conclusions and Future Work

• Software solution for automatic service workflow
  management, which considers not only services,
  but also existing data in workflow construction,
• Ontology describing semantics of the services
  from which the workflows are being composed, as
  well as of the available data and of the users
  which use the software, and of course the domain
  vocabulary,
• Users are able to exchange data and knowledge,
  and cooperate in the workflow construction and
  execution process.
• Future research
• full management of services using the WSRF family
  of standards,
• richer monitoring during the execution phase in
  order to get more relevant QoS parameters