A semantically-enhanced grid registry: Work in progress Sylvia - - PowerPoint PPT Presentation

a semantically enhanced grid registry work in progress
SMART_READER_LITE
LIVE PREVIEW

A semantically-enhanced grid registry: Work in progress Sylvia - - PowerPoint PPT Presentation

A semantically-enhanced grid registry: Work in progress Sylvia Wong, Victor Tan, Weijian Fang, Simon Miles, Luc Moreau School Electronics and Computer Science University of Southampton, UK 1 www.grimoires.org Grimoires Grid Registry Project


slide-1
SLIDE 1

A semantically-enhanced grid registry: Work in progress

Sylvia Wong, Victor Tan, Weijian Fang, Simon Miles, Luc Moreau

School Electronics and Computer Science University of Southampton, UK

www.grimoires.org 1

slide-2
SLIDE 2

Grimoires Grid Registry Project

  • Grimoires

⋄ Grid RegIstry with Metadata Oriented Interface: Robust-

ness, Efficiency, Security

www.grimoires.org 2

slide-3
SLIDE 3

Semantically Discovery of Services

  • Service discovery difficult on the Grid

⋄ Large number of services advertised

  • Semantic descriptions of services

⋄ Filter out most suitable services ⋄ Augment

service descriptions with extra information (metadata) useful to discovery

⋄ Service providers ⊲ Access polices, contract negotiation details ⋄ Users ⊲ Quality of service, reputation metrics

www.grimoires.org 3

slide-4
SLIDE 4

Requirements of Metadata Attachments

  • Annotation to all concepts that influence discovery

⋄ Services ⋄ Operations supported by services ⋄ Input and output types of operations

  • Multiple attachments
  • Third party metadata

⋄ Users to enrich descriptions not foreseen by providers

  • Efficiency in updates

⋄ Some metadata can change frequently, eg user ratings ⋄ Can be updated without republishing the entity or other

metadata attached

www.grimoires.org 4

slide-5
SLIDE 5

Metadata Representation

  • Semantic descriptions as RDF statements

⋄ Subject: entity to be annotated ⋄ Predicate: type of annotation ⋄ Object: value of annotation ⊲ string, URI or RDF graph ⋄ Examples ⊲ (service, mygrid:NumericRating, 8.5) ⊲ (input, mygrid:SemanticType, mygrid:Nucleotide Sequence)

  • Provenance information

⋄ date, author

www.grimoires.org 5

slide-6
SLIDE 6

Metadata Inquiries

  • Multiple query patterns, from simple to complicated

⋄ List of all metadata attached to a service ⋄ List of all entities with metadata that match a list ⋄ RDQL (RDF query language) ⊲ For query patterns not exposed in inquiry interface ⊲ Example: Metadata data values are exact matches currently, use

RDQL to find all services with user ratings > 8.5

www.grimoires.org 6

slide-7
SLIDE 7

Architecture

  • UDDI compatible
  • Multiple web services containers

⋄ Tomcat, Apache Axis, Globus Toolkit 4, OMII

  • Multiple triple store memory backends

⋄ In-memory: Faster, enough for 1 million services ⋄ PostgreSQL, MySQL, Berkeley DB

interface UDDIv2 metadata interface sation module authori− triple Jena store UDDI compliant client Grimoires client uddi4j, etc Registry

www.grimoires.org 7

slide-8
SLIDE 8

Signature-based Authentication

  • UDDI v2 and v3

⋄ Username/password credential scheme ⋄ Authentication tokens

  • Grid environments typically use certificate-based au-

thentication schemes

⋄ Eg, Globus, OMII

  • Grimoires in OMII container

⋄ WS-Security standards compliant SOAP message signing

and verification

⋄ Authentication using Distinguished Name (DN) extracted

from submitted X509 client certificate

⋄ Benefits ⊲ Easy integration into existing Grid security infrastructure ⊲ First step to single sign-on

www.grimoires.org 8

slide-9
SLIDE 9

Performance

  • Preliminary performance tests

⋄ Identify problem areas in efficiency and scalability ⋄ Service publication and inquiry ⋄ Metadata inquiry

www.grimoires.org 9

slide-10
SLIDE 10

Performance Test: Publish

0.1 1 10 100 1000 5000 10000 15000 20000 25000 Time (second) Number of services Publishing 100 Services against Registry Data Size GRIMOIRES, WS GRIMOIRES, BL jUDDI 0.9rc4

⋄ Average publication time: 30ms per service

www.grimoires.org 10

slide-11
SLIDE 11

Performance Test: Inquiry

0.1 1 10 100 5000 10000 15000 20000 25000 Time (second) Number of services Inquiring 100 Services against Registry Data Size GRIMOIRES, WS GRIMOIRES, BL jUDDI 0.9rc4 Inquiry-By-Metadata

⋄ Average inquiry time: 100ms per service

www.grimoires.org 11

slide-12
SLIDE 12

Future work

  • RDFS and OWL support

⋄ Ontology aware service discovery

  • Access control on metadata attachments

⋄ Third-party publication leads to more complicated access

patterns

⊲ Who can annotate a service? ⊲ Who can update a piece of metadata? ⊲ Querying only a subset of metadata?

  • Performance improvements

⋄ Different triple store implementations ⋄ Distributed registry

www.grimoires.org 12

slide-13
SLIDE 13

Summary

  • Presented a semantically-enhanced grid registry

⋄ Metadata interface ⋄ Signature based authentication ⋄ Preliminary performance tests ⋄ Future work

www.grimoires.org 13