Towards Data Management in PeertoPeer Systems: A Perspective of BestPeer

1
Towards Data Management in Peer-to-Peer
SystemsA Perspective of BestPeer

• Aoying Zhou, Weining Qian, Shuigeng Zhou
• Department of Computer Science and Engineering
• Fudan University
• Wee Siong Ng, Beng Chin Ooi, Kian-Lee Tan
• Department of Computer Science
• National University of Singapore

2
Outline

• Project overview
• A brief introduction to BestPeer
• From retrieval to query
• More applications
• On-going work
• Discussion for the challenges

3
Project Overview Applications

• Health care data querying
• Doctors, hospitals store their data separately
• Medicine test or diagnoses may involve complex
  queries
• Genome data processing
• Dozens of databases exist
• Data are changing frequently, since new data are
  added everyday
• Mining tasks are essentially analytical queries
• Data caching in loosely connected organizations
• Data should be shared among different buildings,
  cities or countries
• The system should be scalable to peers and data

4
Project Overview Target

• To build a P2P system for management of
  physically distributed, logically related,
  frequently changing data, satisfied that
• Be scalable to the number of users
• Be scalable to the volume of data
• Support complex services, including information
  retrieval, database query, and analytical query,
  or data mining

5
Project Overview Related Projects

• PIER UCBerkeley
• PIAZZA Washington Univ.
• PEPPER Cornell Univ.
• Pastry Microsoft Rice
• BestPeer NUS Fudan Univ.

6
BestPeer An Overview

• A general purpose peer-to-peer platform
• Key features
• LIGLO server (Location-Independent Global Names
  Lookup Server)
• Self-configurable neighborhood maintenance policy
• Agent-based service mechanism

7
BestPeer LIGLO Server
8
BestPeer Self-configurable Network
9
BestPeer Agent-based Processing

• Share the resource
• Data and information
• Storage power
• Computational power
• Enabling the flexible and customizable complex
  service

10
PeerIS Text Retrieval over BestPeer

• Resource location
• Based on the similarity between peers
• Interests similarity
• Behavior similarity
• More powerful, more important
• From matching to ranking
• Tf-idf-like weighting scheme
• Dictionary is pre-defined

11
PeerDB Querying Databases over BestPeer

• Query on each peer
• SQL queries ? (R, A, C)
• R relation names
• A attribute names
• C conditions
• Approximate search in local databases
• Join the result together
• Moving objects between peers
• Join on-demand
• Mediation vs. warehousing

12
More Applications

• BuddyWeb
• Web caching system
• Peers are grouped together based on the
  similarity of cached content
• PeerOLAP
• Cube are distributed
• Answering analytical queries based on nearest
  cache of cube
• PeerCQ
• Queries are fixed, while data are distributed
• Caching data near the queries

13
What are we doing now?

• Quasi-P QUery AnSwering In Peers
• A unified framework for complex query answering
• Keyword-based search
• Database query
• Analytical query
• Data mining
• Query optimization and view selection
• Data exchange based on XML
• Application virtual lab

14
Challenges

• Routing vs. neighbor maintenance
• Forwarding queries or changing neighbors?
• Performance?
• Security and privacy
• Protect the host from attack which agent is
  safe?
• Different kinds of sharing read-only/read-write/
  
• Multi-granularity services
• Access the multi-granularity data database,
  relation, tuple,
• Collaborate the services provided by different
  peers

15
Discussion

• We have built several prototypes
• Advantages
• Based on flexible BestPeer platform
• Effective and efficient for specific tasks
• Some complex services can be provided
• Problems (and what we are trying to resolve)
• Security is still a problem
• Peers may only work alone but not collaborate
  with others
• What to do next?
• Unified framework
• Providing more complex services

16
Thanks!