1 / 22

An OWL Ontology for QoS

An OWL Ontology for QoS. Glen Dobson (Russell Lock, Ian Sommerville) Lancaster University g.dobson@lancs.ac.uk. Overview. QoSOnt is an OWL ontology for Quality of Service (QoS) I will attempt to answer: What is an ontology? What is OWL? What is QoS? Why is a QoS ontology needed?

jara
Download Presentation

An OWL Ontology for QoS

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. An OWL Ontology for QoS Glen Dobson (Russell Lock, Ian Sommerville) Lancaster University g.dobson@lancs.ac.uk e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  2. Overview • QoSOnt is an OWL ontology for Quality of Service (QoS) • I will attempt to answer: • What is an ontology? • What is OWL? • What is QoS? • Why is a QoS ontology needed? • How should one go about designing such an ontology? • What are the possible approaches? • What are the difficulties? e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  3. What is an ontology? • Standard answer: • “A specification of a conceptualization” (Gruber) • Pragmatically: • A description of the concepts and relationships which exist in some domain using a formal language. • An ontology is an engineering artefact for machine understanding • Its purpose is important. • It should represent shared conceptualisations. • A shared vocabulary is the fundamental component of an ontology • Domain rules are also important e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  4. What is OWL? • OWL is the Web Ontology Language • Supports sharing ontologies via the web • Built on top of RDF (and XML in turn) • Aim is to enable machine “interpretation” of terms and their relationships • It is a Description Logic • Primary constructs are Classes and their Properties • A Class defines a set of Individuals by precisely stating a set of membership conditions. • Main form of inference is subsumption • i.e. is Class B a complete subset of Class A? • + Classification: What Classes is Invidual I a member of? e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  5. OWL in the Semantic Web OWL e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  6. Class Definitions in OWL • Classes can be described • As named resources (as in RDF) • As an enumeration • By constraints on their Properties • By combining other Classes using set operators • Descriptions be combined to give a Class definition using OWL’s: • subClassOf • equivalentClass • disjointWith e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  7. OWL and Inference • A Dog could be asserted to be a Mammal. • Or this classification could be inferred based upon the Class Dog’s Properties (and Property restrictions) • E.g. warm blooded, feeds young with milk, internal fertilisation, etc. • Problem of maintaining a polyhierarchy manually • a Dog is a Mammal, an Animal, a Pet, etc. • Therefore assert a “monohierarchy” and have multiple classifications inferred e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  8. What is your definition of QoS? • Any non-functional aspect of a system that someone may use to judge quality • Extends the definition in distributed multimedia where QoS is primarily concerned with the network (and performance in particular) • In practice we have concentrated primarily on dependability – but the concepts apply beyond this. • What QoS concepts are modelled? • We are primarily concerned with the core concepts of QoS (e.g. attributes, metrics) • Also some consideration to QoS requirements e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  9. Why an ontology for QoS? • To provide a shared vocabulary • For use primarily by machines – but perhaps also in human-readable documents (e.g. requirements documents, SLAs). • To embody machine interpretable “knowledge” • e.g. QoS brokers may need to translate between terms/infer aggregate values/convert units, etc. • Also the provision of QoS description and reasoning capabilities to the semantic web e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  10. QoS Sub-Systems Service Payment Banking systems Service Discovery Service Differentiation Service negotiation Service Agreements Service Operation Service Monitoring Service Mediation QoS Prediction Re-negotiation Workflow Planning Law e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  11. What “added value” could a QoS ontology provide? • Translation based upon machine “understanding” • Translation of units, computation of composite metrics, inference of aggregate QoS for workflows • Leeway in syntax matching • i.e. multiples terms can refer to the same thing • An interlingua for translation between other QoS languages • A means for agents to communicate e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  12. QoSOnt Structure • At the core of QoSOnt is a taxonomy of Attributes and Metrics • i.e. two trees formed using the subClassOf construct • An attribute is e.g. reliability, performance • A metric is e.g. Probability of Failure on Demand, Transactional Throughput • This becomes a (complex) directed graph once properties are considered • e.g. The Property hasMetric (and its inverse isMetricOf) is the basic link between the attribute and metric trees e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  13. Danger of Ontology Creep • Should we provide a model to represent: • Time • Currently we do – but we should instead use the OWL-Time ontology. • Ways of composing metrics, Mathematical constructs that don’t exist in OWL • This originally put us off and thus we have a separate XML language as well as the ontology. • Ways of composing services • We currently use a very shallow model – but perhaps this is all that is needed? e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  14. QoSOnt High-Level Structure Metric Instances Metric Layer Metrics Attribute Layer Performance Dependability Etc …. Low level concepts Base concepts Time Underlying OWL e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  15. IFIP 10.4 Dependability Taxonomy • Our example of an attribute layer ontology • Familiar Fault-Error-Failure model • Main point of linkage is DependabilityAttribute is a subclass of QoSAttribute • Shows how a detailed model of certain attributes can help • E.g. without the definition of Failure, Failure Domain it is impossible to be specific about what a Probability of Failure On Demand refers to e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  16. Overview of Metric Definition e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  17. Representing QoS Requirements • As OWL Classes using built-in OWL constructs • Datatype support is poor • No consistent way of using custom XML types • Reasoning support for quantification over datatypes (e.g. allValuesFrom 0-100) is poor. • Level of datatype support mandated by OWL spec is poor • Using QoSOnt defined Classes, Properties, Restrictions, etc. • As a separate (XML) language referencing the ontology e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  18. SQRM Tool e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  19. Requirements Matching in SQRM e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  20. Evaluation/Future (1) • An ontology is a good idea – but a large-scale standardisation effort is required • Need external input in order to evolve • Two interested parties are now involved • Requirements representation and matching using built-in OWL features would be nice • Need to wait for OWL to develop • Need to look at SWRL (Semantic Web Rules Language) • E.g. would provide a neater way to express unit conversions e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  21. Evaluation/Future (2) • Need to work on tools that make use of QoSOnt (and also enhance SQRM) • Difficult to evaluate otherwise since the purpose is machine-machine understanding • But are there really a lot of QoS “semantics” to model? • Service Composition/Workflow • Integrating existing work with ontology e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

  22. Questions For more information: http://digs.sourceforge.net e-Science AHM, Nottingham. September 20th, 2005. Glen Dobson: g.dobson@lancs.ac.uk

More Related