ICONS CERIA Contribution Outline

1
ICONS CERIA ContributionOutline

• U. Paris 9 Dauphine
• Witold Litwin

2
Project Goals

• Main (in terms of man power)
• WP5 Distributed Content Repository
• Distribution of ICONS Processes and Structures
• Contribution to Access Algorithms and Data
  Structures (ICS is the leader)
• Design of System Architecture for Integration of
  Pre-existing Heterogeneous Information Sources
• Contributions to WP1-WP8

3
Project Deliverables

• Input to other teams in charge of WPi
• Milestone Technical Reports
• Etc (see the contract)

4
Project Stuff

• 1 Senior Researchers
• 2 Researchers (Prof. Ph. D. student)
• unlisted participants
• junior students

5
Technical basis of our input

• Scalable Distributed Data Structures (SDDSs)
• technology now also vaguely called NOW, P2P, Grid
  Computing
• Theory
• Various algorithms for scalable data partitioning
• hash LH schemes
• range RP schemes
• k-d kRP
• high-availability LHrs
• SDDS-2000 prototype

more at ceria.dauphine.fr
6
Research Directions to be investigated for ICONS

• Theoretical identification of schemes suitable
  for project goal
• Study of SDDS-2000 as storage and distributed
  processing service for ICONSs upper levels
• including on the Web
• Globus services ?
• RAM oriented
• including RAM DBMS servers/mediators (AMOS-2,
  Timesten)

more at ceria.dauphine.fr
7
Technical basis of our input

• Multidatabase System Design
• Multidatabase Architecture
• Multidatabase Language
• MSQL
• OSQL
• object oriented (with Chomicki)

more at ceria.dauphine.fr
8
Research Directions to be investigated for ICONS

• Multidatabase Architecture for Knowledge
  Management
• general, Mariposa, OO
• Application of Multidatabase Language Features to
  XML
• known MSQL functions
• elementary queries
• multiple queries (higher-order predicate
  calculus)
• MDB bibliographic systems now use such functions
  widely
• meta-search engines also
• New functions ?
• Implementation Issues
• multi-relational algebra (with Grant, Sellis and
  Roussopoulos)
• Objectlog (with T. Risch)
• OSQL
• OO-specific MDB mappings

more at ceria.dauphine.fr
9
Concurrent Analysis (why these technics ?)

• New technology
• mature at research level
• made in Europe
• although patents are US (HPL IBM)
• Established practical interest
• Buzz around concepts of scalability, P2P, mdb
  access, high-availability, distributed RAM
  processing
• Buzz around Trouble Ahead (Risk factors)
• Research may succeed or (partly) fail

10
ICONS Specific Needs

• No known work on Scalable Distributed Knowledge
  Management
• Interactions with other teams
• Research teams
• Development folks
• ICONS application of SDDS-2000
• Freely available at CERIA site
• had dozens of downloads without major complaints
• except for known limitations
• no duplicates, record size lt 8K
• Can be autonomously applied by partners

11
Our Rules of Thumb

• Simple Things are Complex
• especially to realize
• Complex Things are Unfeasible
• idem
• Plan Less to be More Likely to Deliver
• not the other way around
• You Plan Only What You Know About
• Your Time Gets Unpredictably Eaten by what you do
  not know while planning