Open Grid Service Architecture Data Access

1
Open Grid Service Architecture Data Access
Integration(OGSA-DAI)

• University of Edinburgh
• National e-Science Center
• IBM, United Kingdom
• University of Manchester
• Oracle, United Kingdom
• University of Newcastle

2
Whats the OGSA in OGSA-DAI ?

• OGSA-DAI is a middleware extension of the Open
  Grid Service Architecture (OGSA)
• OGSA is a framework that,
• Defines standard mechanisms for creating, naming
  discovering Grid services
• Provides location transparency protocol
  bindings for service instances
• Supports integration with underlying native
  platforms

3
OGSA Overview
4
Data Access Integration (DAI)

• Objective is to achieve seamless sharing of data
• As introduced in Liao, K., and McLeod, D. (2000)
  Federating Neuroscience Databases, DAI comprises
  of,
• Information Discovery
• Finding out what information can be shared in the
  first place
• System Level Interconnection
• Collection of technologies and mechanisms that
  facilitate the actual transmission of data
• Semantic Heterogeneity Resolution
• Making sense of the data once it is found and
  accessed

5
OGSA-DAIs Focus

• Information Discovery
• System Level Interconnection
• In summary, to provide the basic primitives to
  construct higher level services that allow data
  federation distributed queries to take place in
  a Virtual Organization

6
OGSA-DAIs Architecture

• Comprises of 3 Grid services
• Grid Data Service Registry (GDSR)
• Keeps a record of data sources in the Grid,
  their respective exposed schemas and functions
  available for use (static as of now)
• Grid Data Service Factory (GDSF)
• Contacts the desired remote data source,
  authenticates the client and establishes a GDS to
  link the client directly with the remote data
  source
• Grid Data Service (GDS)
• Acts as an interface via which the client sends
  queries directly to the remote data source

7
OGSA-DAI Operational Overview
8
Query Execution

• Based upon exposed schema information in the
  GDSR, query is formulated and sent to remote data
  source
• Allows functions,procedures or features written
  specifically for the remote database to be
  utilized, hence minimizing the loss of vital data
  as a result of the sharing process

9
Result Set Delivery

• Dependent on size of result set
• Small sets of data can utilize copy methods
• Large sets of data require greater complexity
• One possibility is to copy results to an
  edge-server close to client Initiates
  communication between prior GDS to data source
  and new GDS to where results are sinked
• Another alternative is to use URIs in the event
  the sink is outside the Grid

10
In Progress Future Work

• Design and implementation of the OGSA-DAI
  Distributed Query Processing (DQP) Grid Service
• Choosing a suitable data replica based on some
  criteria, for example, closestor least
  expensive
• Parallel downloads from data resource replicas or
  copies to improve performance

11
References

• Liao, K., and McLeod, D. (2000) Federating
  Neuroscience Databases, University of Southern
  California
• Hume, A., and Atkinson, M. (2004) Brief
  Introduction to OGSA-DAI, E-Science Project,
  United Kingdom
• Wohrer, A., Brezany, P., Janciak, I., and Tjoa,
  A. (2004) Virtualizing Heterogeneous Data Sources
  on the Grid Design Concepts and Implementation,
  University of Vienna