Applying Grid Technologies to Distributed Data Mining

1
Enabling Access to Federated Grid Databases An
OGSA-DAI ODBC Driver
Michael J. Jackson1 Ashley D. Lloyd2 Terence M.
Sloan1
1EPCC 2Curtin Business School Edinburgh
University Management School
2
Overview

• Why develop an OGSA-DAI ODBC driver?
• ODBC
• OGSA-DAI
• Design and Development
• What does an OGSA-DAI ODBC driver give us?
• Issues and Concerns

3
Why?

• Facilitate use of standard data analysis tools in
  a Grid environment
• Remove need for Grid awareness
• Allow use of existing data analysis skills in a
  Grid environment
• Improve rate of adoption of Grid technologies
• Data analysis tools
• SPSS, SAS
• How can standard data analysis tools access
  Grid-enabled databases?
• An ODBC driver for OGSA-DAI

4
Open DataBase Connectivity ODBC
ODBC data source
Database API
ODBC API
ODBC API
Data source name
Reside on same host
5
ODBC Advantages

• Application developers
• Applications can be database-independent
• No need to compile against database-specific
  libraries
• Call-level interface execute SQL generated at
  run-time
• Change a database gt only change driver and
  configuration
• Database manufacturers
• An ODBC-compliant driver allows the database to
  be a back end for any ODBC-compliant application

6
OGSA-DAI

• Open Grid Services Architecture Data Access and
  Integration
• Extensible framework for data access and
  integration
• Expose heterogeneous data resources to a Grid
  through web services
• Data operations
• Access, update, management and integration
  relational, XML, files
• Compression and transformation
• Delivery to URLs, FTP, GridFTP, mail, other
  services
• Base for developing higher-level services
• Data federation and distributed query processing
• Data mining
• Data visualisation

7
Accessing Data Resources via OGSA-DAI
OGSA-DAI Perform document
JDBC API
OGSA-DAI Response document
8
An ODBC Driver for OGSA-DAI
ODBC API
ODBC API
Data source name
OGSA-DAI Response document
OGSA-DAI Perform document
JDBC API
9
A Simple Scenario

• Data analysis
• ODBC view
• Connect to OGSA-DAI ODBC data source
• Submit a SELECT FROM table query
• Get back the results
• Disconnect from the data source
• OGSA-DAI view
• Connect to an OGSA-DAI data service
• Construct a Perform document holding the query
• Send it to the service
• Receive a Response document from the service
• Parse it to get the results

10
Development Options

• Implement an OGSA-DAI ODBC driver
• From scratch
• Use an open source ODBC driver
• Extract a data resource-independent skeleton
• Customise it to OGSA-DAI
• Use an ODBC SDK
• OpenAccess
• Simba
• Syware

11
Using an SDK

• Proof of concept
• Prototype within a tight time-scale
• OpenAccess SDK
• 30 day evaluation licence
• Provides an ODBC driver
• Developer codes an Interface Provider (IP)
• Supports Java development gt exploit OGSA-DAIs
  client toolkit

12
An ODBC Driver for OGSA-DAI using OpenAccess
OGSA-DAI Perform document
OGSA-DAI CTk API
OpenAccess API
OGSA-DAI Response document
Data resource configuration (e.g. service URL)
13
Testing

• OpenAccess ODBC SQL query tool
• Submit SQL statements to an ODBC data source
• Present the results
• EPCC
• OGSA-DAI ODBC data source on a PC
• ODBC driver OGSA-DAI service URL
• Curtin Business School
• OGSA-DAI server and services
• Database

14
What does this give us?

• Transparency
• Database location
• Changes are restricted to the OGSA-DAI server
• Client applications are unaffected
• Database product
• Global access of data
• Publish service URL
• Security
• Database user names and passwords reside on
  OGSA-DAI server
• Clients can be required to provide credentials to
  connect to OGSA-DAI services

15
Data Federation
ODBC API
ODBC API
Data source name
OGSA-DAI documents
Virtual database
16
Issues and Concerns

• OGSA-DAI WSI / WSRF compliance
• Prototype developed using OGSA-DAI OGSI
• Data source includes OGSA-DAI factory service URL
• OGSA-DAI WSI or WSRF data service URL
  resource ID
• Driver development
• Complete the OpenAccess IP
• Write a pure OGSA-DAI ODBC driver from scratch
• ODBC conformance
• Cursors, sessions, transactions, timeouts,
  meta-data
• Analysis of SAS or SPSS ODBC usage
• Efficiency

17
Questions