Mediation und Indexing in P2P-basierten Informationssystemen

1
Mediation and Indexing in P2P-based Information
Systems
Super-Peer-based Routing in the Edutella-Network
2
Overview

• Peer-to-Peer Topologies
• Pure p2p
• Super peer networks
• Super-Peers in Edutella P2P
• Query Routing and Processing
• Creating the p2p network
• Relaying queries and results
• Future
• Advanced super-peer topologies

3
Peer-to-Peer Networks
P2
P4
P3
P5
P9
P8
P6
P1
P7
4
Hybrid Networks
Files C,D,E
P2
P4
Files B,C,D
P3
P5
Index
P6
Search A
File B
P1
P7
File A
5
Super-Peer Networks
Files C,D,E
P2
P4
Files B,C,D
P3
A SP3 BCD SP2
P5
SP1
SP2
Search A
P6
SP3
File B
P1
P7
File A
6
Edutella P2P Network

• 2 Roles
• Information providing role (Provider)
• Information consuming role (Consumer)
• Implemented in JAVA
• Based on JXTA, Sun P2P Architecture
• Edutella Common Data and Exchange Model
• Java-Model for representing Queries
• RDF based Query Language (RDF-QEL-i)
• Different levels of expressiveness

7
RDF Resource Description Framework

• RDF Statements
• Subject Predicate Object
• RDF Schema
• DC Dublin Core, LOM Learning Object Metadata
• dctitle -gt http//purl.org/dc/elements/1.1/title
  
• Example Database
• ai.html dctitle Artificial Intelligence
  
• ai.html dctype Book
• ai.html lomLanguage de

8
RDF Graph
Query
Statement
RDF-QEL-3
9
Old Peer Design
Provider Peer

• Provider Peer
• JXTA Pipes for Communication
• Edutella Query Service
• Connection to Metadata Repository

Metadata Repository
Database Backend
Query Service
JXTA
10
New Peer Design

• Services
• New Roles
• Query routing, indexing, provider-peer
  registration
• Topology
• Connections between super-peers
• HyperCup, Broadcast-Network
• Events
• Component-based architecture
• Modules for communication, processing, indexing,
  management, statistics
• Subsystems send and receive Messages (like
  Java-AWT,Swing)

11
Super-Peer Implementation
12
Peers joining the Network
P3
Binding Service
ai.html dctitle Artificial Intelligence ai.
html dctype Book ai.html lomLanguage de
P4
SP1
LOM P3,P4 dctitle P3 dctype P3
Indexer
SP/P Routing Index
13
BIND - Binding Service

• Provider-peers joining and leaving the super-peer
  network
• Provider-peer
• Try to find a suitable super-peer
• Create a description of the providers
  capabilities
• Send messages to a super-peer
• Monitor the state of the super-peer
• Super-peer
• Maintain a database of all connected
  provider-peers
• Notify the Indexer to update routing indices

14
Indexer

• Linked to the Binding Service
• Routing indices based on provider descriptions

Element
Peers
Schema
P3, P4
LOM
Property
dctitle
P3
Property/Range
Classification AI Books
?
15
Constructing SP/SP Index
P3
Binding Service
LOM P3,P4 dctitle P3 dctype P3
P4
SP2
SP1
SP/P Routing Index
DC SP1
Topology
empty
SP/SP Routing Index
C1
16
Routing Mechanisms
DC P3,P4 lomLanguage P3
SP/P Index
Router
P3
Routing Table
Routing Service
P4
Processor
Query
SP2
dctitle
Routing Table
SP/SP Index
Router
DC SP2 LOM SP4
17
ROUTE - Routing Service

• Route Queries received from a consumer-peer to
  connected super-peers and providers
• Routing between super-peers based on SP/SP index
• Routing to provider-peers based on SP/P index
• Characterize Query
• Relaying of results

18
QEL - Edutella Query Service

• Process Queries received from consumer-peers
• Provider-Peer
• Resolve using database backend
• Re-used code from original Edutella provider peer
• Super-Peer
• Accept queries from consumers and
• Forward queries to routing service

19
Future

• Dynamic Indices
• Monitoring traffic and query characteristics
• Semantic Clustering
• Group peers with similar metadata
• Peer Trading
• Move provider-peers to other super-peers
• Advanced Super-Peer Topologies
• HyperCup
• Mediation
• Translating queries between different RDF schemes

20
Mediation

• Translate between different RDF schemes
• Similar attributes in multiple schemes
• Global RDF schema
• globalattribute -gt peerattribute

LOM
P3
lomIdentifier
Mediator
Query
DC
globaltitle
P4
dctitle
21
Conclusion

• Efficient Query Routing
• Routing indices for query routing
• Few messages for synchronization
• Extendable Peer Design
• Event passing interface
• Modular concept