An Ontology-based Approach for Managing Software Components

1
An Ontology-based Approach for Managing Software
Components

• by
• Vladimir Kolovski

2
Motivation

• Using off-the-shelf components has proven
  invaluable
• development time/production costs dramatically
  decrease
• usability increases
• Application servers as component-based middleware
• platform for deploying components
• dynamic loading, naming services, load balancing,
  etc.

3
Motivation Use Cases

• Managing components still a cumbersome task
• Component dependencies and versioning
• Licensing
• Service Classification and Discovery
• XML configuration files not enough
• Explicit formal model would help

4
Introduction to Ontologies

• Ontology is a controlled vocabulary that
  describes objects and the relations between them
  in a formal way
• May have different structure
• Logic-based semantics
• Need to choose a representation language

5
OWL

• Web Ontology Language (OWL) is a W3C
  Recommendation
• Based on XML syntax
• OWL Lite, OWL-DL, OWL-Full
• OWL-DL decidable subset of First Order Logic
• Classes, Individuals, Properties, Restrictions

6
Component Ontology

• Profile
• Capability descriptions
• serviceName, serviceParameter, contactInformation
• Implementation
• dependsOn, conflictsWith
• Grounding
• Mapping between syntactic and semantic metadata

7
Java Management Extensions

• Open technology for component management
• Defines interfaces of managed beans
• MBeans hosted by an MBeanServer which allows
  their run time deployment and manipulation
• JMX is just a specification, MX4J used as an
  implementation for this project

8
Implementation

• MBeanServer scans for deployed / undeployed
  components
• Ontology and reasoner deployed as an MBean
• Use OWL API for parsing and Pellet for querying

9
Implementation

• Component has its own metadata (manually entered
  for now)
• Every component is an instance with properties
• Query ontology when a component is added, look
  for conflicting ones
• Also query ontology when a component is removed

10
How do we Query an Ontology?

• OWL is based on top of RDF
• RDF model is a graph of triples (Resource,
  Property, Value)
• Express the component ontology as an RDF graph
• Use RDQL, SQL-Like language

11
RDQL Example

• SELECT
• WHERE (?x, CompconflictsWith, compcomponent1)
• USING
• Comp FOR lthttp//www.mindswap.org/dav/ontologi
  es/Comp.owlgt
• Returns a list of individuals conflicting with
  component1

12
Future Work

• Short Term
• Integrate with JBoss Component Loader so it
  would work with MBean dependencies
• Evaluation (Pet Store)
• Licensing
• Long Term
• Automate retrieving metadata from configuration
  files
• Component grounding

13
Related Work

• JBoss
• JBossMX implementation of JMX
• MBean dependencies
• RPM
• Package management system
• KaON (JBossMX, focused on run-time support, not
  using OWL)
• Model Driven Architectures
• Lack logical formality to query app server at
  runtime

14
Questions Comments