Peer-to-Peer Data Integration Using Distributed Bridges

1
Peer-to-Peer Data Integration Using Distributed
Bridges
Candidate Thesis for M. A. Sc. in Electrical
Engineering

• Neal Arthorne
• B. Eng. Computer Systems (2002)
• Supervisor Babak Esfandiari
• April 12, 2005

2
Introduction

• Multiple autonomous heterogeneous data sources
• E.g. chemistry and genetics databases, digital
  repositories, astronomy databases
• Data is distributed and network-accessible
• Each data source may use a different syntax or
  query language (SQL, Web Services etc.)

3
Related Work

• Federated database systems Sheth, 1990
• Global federated schema
• Mediator approach Wiederhold, 1992
• Databases wrapped in a software layer that
  translates to a common information model
• Middleware lies between user applications and
  data sources
• Theoretical description of data integration
• Schemas mapped with FOL statements LAV or GAV
  approach

4
Related Work Contd

• OWL/RDF/RDFS used to describe semantic
  relationships ontologies
• Peer-to-Peer Data Integration WWW approach to
  integrating data
• PIAZZA (Halevy et al.), Lenzerini, Franconi
• Focused on query optimization and decidability in
  FOL systems

5
Limitations of Related Work

• Global shared schemas are fragile and not
  scalable
• Centrally located and administered
• Changes affect all component databases or
  middleware
• P2P data integration is limited
• Semantic differences not addressed
• Centrally stored mappings
• Large databases not compatible with centralized
  metadata

6
Proposed Solutions

• User-contributed mappings between schemas
  (bridges)
• Fully de-centralized distribution of mappings
• Anyone can publish a new mapping
• No global schema means improved scalability
• Provide semantic mappings for data
• Distributed searching compatible with large
  databases
• Use existing Universal Peer-to-Peer (U-P2P)
  framework

7
Universal Peer-to-Peer (U-P2P)

• Peers share XML metadata with binary attachments
• Communities formed around a shared XML Schema
• Community itself is published anyone can create
  a community
• Flexible deployment pluggable Network Adapters

Book Comunity
0..
1
Resource
1
0..
lt?xml version1.0?gtltbookgt lttitlegtWar
Peacelt/titlegt lte-textgtfile//...lt/e-textgt
Attachments
8
P2P Data Integration with U-P2P

• Proposed Bridge Community and bridge schema
• Anyone can publish a bridge
• Includes simple semantic relation
• Attached mappings and/or transforms
• U-P2P modularized for database proxies
• Distributed Network Adapter (Gnutella)
• Compatible with large databases
• No central indexing servers

9
Bridges in U-P2P
Resource
Community
Bridge
10
Example Bridge
ltbridgegt lttitlegtDSpace to Fedora
bridgelt/titlegt ltdescriptiongtlt/descriptiongt
ltbridgeMappinggt ltsourcegt
ltcommunitygtd2a9d6f78dcf91828f68a52f78260e05lt/commu
nitygt ltresourcegt134d8f8ecd57acb35206b4cd13e3
8622lt/resourcegt lt/sourcegt
ltrelationgtowlsameAslt/relationgt lttargetgt
ltcommunitygtd2a9d6f78dcf91828f68a52f78260e05lt/comm
unitygt ltresourcegtda1058314b7d8890fc7df7f879a
0a7dblt/resourcegt lt/targetgt
lt/bridgeMappinggt lttransformListgt
lttransformgt ltfilegtfile//lt/filegt
lt/transformgt
11
Case Study Digital Repositories
Peer A
DSpace Community
Generic Central Server
Fedora Community
Peer C
Fedora Database
DSpace Community
Peer B
Fedora Community
Fedora Community
Proxy
Gnutella Protocol
Centralized P2P
12
Conclusion

• P2P approach to integration anyone can create a
  bridge
• Fully-distributed network adapter brings in large
  data sources via proxies
• Demonstrated integration with digital
  repositories
• Simple semantic relationship (OWL)
• Query translation

13
Future Work

• Manual navigation between schemas
• Need to automate retrieving bridges
• XPath query translation is limited
• Need to provide robust query translation modules
• Translate instance data
• Semantic relationships not exploited
• Use OWL ontologies to give bridges a context
• Software agents can be introduced to discover and
  use bridges