Bez%20tytulu%20slajdu

1
COMPUTER ENGINEERING DEPARTMENT
TECHNICAL UNIVERSITY OF LODZ, POLAND
Wrapping Relational Data Structures to
Object-Oriented Databases in the Data Grid
Architecture
Radoslaw Adamus, Kamil Kuliberda, Jacek Wislicki,
Kazimierz Subieta
SOFSEM-SRF 2006, January 21-26, Merin, Czech
Republic
2
Wrappers Introduction

• problem started in late 80-ties first
  approaches concerning object-oriented wrappers
  built on top of the relational database
• motivation for the wrapper business data
  virtual integration
• transparency grid needs this
• data optimization utilization of optimization
  mechanisms in both data models
• used techologies Stack-Based Approach (SBA),
  M0 model, SBQL, Updateable Views
• Data Grid conception
• integrates business information
• data-intensive grid simply data grid
• simplification of data access
• common grid data model for virtualization of
  local schemata canonical data model
• implementation is based on ODRA database engine

3
Data Grid architecture
4
Object-to-relational wrapper

• it combines exploiting SBQL query engine and SQL
  query engine
• externally data structure is designed according
  to the object-oriented model
• and business intention of global schema
  front-end
• internally relational structures are presented
  as M0 model back-end
• updateable object views define mapping between
  front-end and
• back-end, B into F for querying, F onto B for
  updating
• the wrapper utilizes three major optimization
  methods for back-end side
• - rewriting (for instance pushing selections
  before joins),
• - indices (i.e. internal auxiliary structures
  for fast access),
• - fast joins (e.g. hash joins)
• during optimization, SBQL syntactic query tree
  is first modified from business to M0 model
• wrapper obtain much larger tree, but addressing
  a primitive object database that is 11 mapped
• to corresponding relational database

5
Wrapper architecture
6
Conclusions, future work

• presented wrapper architecture enables building
  generic solutions allowing presentation of data
  stored in various relational resources as
  object-oriented models
• these models are visible at the top level of the
  grid and accessing the data with object query
  language
• presented optimization process assumes correct
  relational-to-object model transformation (with
  no loss of database logic)
• allow on accessibility of the relational model
  optimization information such as indices and/or
  primary-foreign key relations
• the method is currently being implemented as a
  part of our new project ODRA devoted to Web and
  grid applications
• currently, based on presented conception, our
  research concerns object-to-semistructured
  wrapper exploiting native optimization mechanisms
  for querying semistructured data (e.g. Lore DBMS
  XML-based)

7
Thank you for your attention!