Transcript The Product Class - Mathematical & Computer Sciences

A Virtual Distributed Database Model
for Creating a Database Federation
Jaspreet Singh Pahwa
PhD Year 2
School of Computer Science
Cardiff University
Overview
•
•
•
•
•
Motivation
Hypothesis
Product Class
Product Class Database
Grid Enabled Multiple Database Search
Service (MDSS)
• MDSS Federation
• Conclusions and further work
Motivation
• Information held by large number of data sources in a
Wide Area Network.
• Access to such information held by large number of
organisations.
• Information processing with aid of independent
mechanisms, and
• Sharing of information with those who need it.
Hypothesis
• Enabling product definition infrastructure.
• Creating standard product definition using the infrastructure.
• Subscription of product definitions and product creation by
product suppliers.
• Building a virtual distributed database using the Grid
infrastructure.
• Participation of product suppliers in the virtual distributed
database as part of a federation.
• Advertisement of products in the virtual distributed database by
Product Suppliers.
• Product search using product definition and the Grid.
The Product Class
• A standard product definition for providing
product information.
• Acts as a template and provides meta
information for creating actual product
information.
• Is made up of a number of specifications
conforming to different specification types.
• Can be used by Product Suppliers for product
advertisement in the federation.
The Product Class and its various
specification types
Product Class
Specification
Value
Table Specification
List of Column
Specifications
where each
column
specification has
list of values
relative to other
column
specification
values.
List Specification
Specification Group
Specification
Value
Table Spec
Value
List Specification
…
Sub-Specification
Group
Product Class
Specification
Sub Product Class
Specification
Product Class Versioning
• Versioning of Product Class/Specifications.
• New products emerge.
• Existing products evolve and are assigned
more attributes.
• New versions created and assigned.
• Product suppliers notified.
• New classes downloaded.
Product Class Subscription
Transport Layer
Product Class
Database
Table
Table
Product Supplier
Database
<xml>
…
…
…
</xml>
Table
Xml
Parse
Service
Table
…
…
Transport Layer
The Product Class Database
• A tool for creating Product Classes and
Product Class Specifications.
• Enables Product Class versioning.
• Releases Product Class for subscription
by Product Suppliers.
• Provides meta information for building
product search criteria.
Product Class
Category
Product Category
Category Hierarchy
Specification
Product Class Version
Product Class Definition
List Version
Specification Group
Version
List Specification
Specification Group
List Definition
Specification Group
Definition
Table Version
Table
Specification
Sub Product Class
Version
Sub Product Class
Table Definition
Product Class Database Structure
Sub Product Class
Definition
The MDSS
• Multiple Database Search Service.
• A database centric Grid service for accessing
and processing information from large
number of data sources.
• Searches supplier data sources based on a
search criteria.
• A channel for product suppliers for advertising
products to a large user base.
• Built using Globus Toolkit 3.0.2 (Core).
The Search
• What to search?
- a search criteria submitted by the user.
• Where to search?
- data sources that match user’s request.
• What are the available resources to propagate the
search?
- the available Database Search Services in
the MDSS
Search Criteria
Where to Search?
<NewDataSet>
<Supplier>
<IDSupplier>1</IDSupplier>
<SupplierWS>http://131.251.42.40/test/categories.asmx</SupplierWS>
</Supplier>
…
</NewDataSet>
Available GSHs
<NewDataSet>
<GridServiceHandle>
<IDGsh>1</IDGsh>
<GSH>http://131.251.42.40:8080/ogsa/services/DatabaseSearchService</GSH>
</GridServiceHandle>
<GridServiceHandle>
<IDGsh>2</IDGsh>
<GSH>http://131.251.47.110:18080/ogsa/services/DatabaseSearchService</GSH>
</GridServiceHandle>
…
</NewDataSet>
The Architecture
Two major components:
• Master Grid Service: identify resources available,
register new resources, job allocation, data parsing and
data aggregation.
• Database Search Service: data search, data
aggregation, data parsing, collaborative working.
Single user
VO
Grid Engine
Metadata
Product Classes
Portal
(Search Criteria Specified here)
* Metadata Query
Master Grid Service
*Job Allocation Grid Service * Collation of datasets
(based on the OGSA Model)
Grid DB Search Service
Grid DB Search Service
Grid DB Search Service
Apache Axis Soap Server
<<soap messages >>
WS Layer
WS Layer
<<soap messages >>
WS Layer
WS Layer
Supplier databases
- MDSS Architecture -
<<soap messages >>
WS Layer
WS Layer
Design Features
• Ability to register new data sources.
• Ability to enhance MDSS capability by adding more
machines in the Grid cluster.
• Supplier databases free of Grid Complexities.
• Ability to identify data sources to be searched in real
time.
• Distribute search equally among available GSHs.
• New instance of Database Search Service for each
operation.
Job
Execution
Database
Search
Service
Grid
Service
Manager
GSH
Document
Parser
Master
Grid
Service
Supplier
Document
Parser
Data
Aggregation
Xml
Converter
OGSA Container
Master Grid Service Modules
MDSS Federation
• Information retrieval model for retrieving product
information without the need of creating federated
schema.
• Encompasses the MDSS System and all participating
SD Systems.
• Cooperation to the centralised MDSS System for
advertisement of products to VOs.
• Provides an environment for data models to compete
with each other.
• Database access mechanism Grid enabled and not
data sources.
• Component data models participate in the federation
by providing web service interface.
Single user / VO
The MDSS System
Database Search
Job Allocation
Collation of Datasets
Web Service Layer
Web Service Layer
Web Service Layer
Component Schema
(Product Class)
Component Schema
(Product Class)
Component Schema
(Product Class)
Component DBS
Component DBS
Component DBS
Supplier 1
Supplier 2
Supplier n
The MDSS Federation
Design Features
• Component DBS retain local autonomy.
• Schema integration or creation of federated
or external schemas not required.
• Uniform data model addresses the issue of
semantic heterogeneity.
• Supports different kinds of autonomies.
Conclusions & Further Work
• MDSS Federation – a Grid enabled virtual
distributed database model for product
advertisement.
• Creation of downloadable product classes.
• Product class subscription.
• Multiple instances of Master Grid Services.
• Data Streaming.
• Memory Resident databases.
Questions?