Database Access and Integration Workshop 1516 April 2002, NeSC Report Back

1
Database Access and Integration Workshop15-16
April 2002, NeSC(Report Back)

• Jasmin Wason
• Zhuoan Jiao
• DAI Workshop Online Documents
• http//umbriel.dcs.gla.ac.uk/NeSC/general/esi/past

2
Introduction to Database Task Force Norman
Paton (5 mins)

• AIMS
• Identify requirements of DB and information
  management applications within a Grid setting.
• Develop road map for RD of DB functionalities in
  a Grid setting.
• Standards for Grid/DB, links through GGF (wider
  community).
• Advise Architecture Task Force on requirements
  for DB support.
• Foster reference implementations.
• ACTIONS
• OGSA-DAI Project funded (18 months)
• GGF Working Group
• See www.cs.man.ac.uk/grid-db

3
Database requirement analysis Dave Pearson (45
mins)

• Key requirements
• Represent complex data models
• Maintain metadata
• Technical, contextual, mappings and rules
• Establish reliability and quality of data
  (provenance)
• Make data more accessible (publishing
  discovery)
• Restrict who can read modify data (access
  control)
• Get data in a ready-to-use state
• Data retrieval
• locational transparency, everyday language
• Analysis and interpretation capability (e.g.
  evidential reasoning)
• Managing large volumes of data

4
DAI Workshop Show-and-TellAstroGrid Project
(Clive Page, Leicester University)

• Belfast, Cambridge, Edinburgh, Leicester,
  Manchester, RAL, UCL
• Main aims
• Federate the worlds distributed astronomical
  data.
• Data mining using Data Grid and Web Services
  concepts.
• Design virtual observatory based on OGSA
• Database Challenges
• Interoperability XML standards for queries,
  results metadata.
• Resource discovery UDDI not very suitable
  (business-centric)
• Query language SQL poorly matched to
  astronomical needs.
• Indexing celestial position, poorly supported
  now.
• Handling complex data structures. OODBMS a
  failure?

5
DAI Workshop Show-and-TellGEODISE Project
(Zhuoan Jiao, Southampton University)

• GEODISE - Grid Enabled Optimisation and Design
  Search for Engineering
• Goals To bring together and further the
  technologies of Design Optimisation, CFD, GRID
  computation, Knowledge Management Ontology in a
  demonstration of solutions to a challenging
  industrial problem.
• Partners Southampton University, Oxford
  University, Manchester University, RAE Syetms,
  Rolls Royce, Fluent,
• Current Status
• Simple ontology for describing optimisation
  process
• CFD and optimisation code wrapped as web services
• Computations run on Condor web service
• Automatic web service message logging, user
  information logging
• Web based queries to databases
• Visualization of the data using SVG web service

6
DAI Workshop Show-and-TellGEODISE Project
(continue)

• Future Directions (mainly regarding database work
  package)
• Wrap databases as web/Grid services
• Query interface to distributed databases
  (possibly ontology based)
• Knowledge assistance in optimisation and design
  decision making
• Data mining
• Evolution of database schema to cope with WSDL
  changes
• Investigate native XML DBMS as alternative to
  relational
• Work flow management (WSFL, etc.)
• Distributed file access (GridFTP)
• Security

7
DAI Workshop Show-and-TellDistributed Query on
the Grid (Paul Watson, Newcastle University)

• Projects
• Polar (parallel query processing over the Grid)
• Database Access and Integration (DAI) for the
  Grid
• MyGrid
• Research Issues
• Schema integration
• Distributed query execution infrastructure
• e.g. how to allocate machines to perform the
  join queries
• Adapting to changing information
• Adapting to failures
• Workflow

8
DAI Workshop Show-and-TellXSPAN Project A
Cross-Species Anatomy Network
(Albert Burger, Heriot-Watt University)

• Will allow anatomies from different species to be
  compared.
• Links between textual ontology and 3D anatomy.
• XSPAN Core
• Generic XML schema for embryo anatomy (for
  interoperability)
• DBs have different schemas but map data to
  generic XML Schema.
• Graphical user interface and program level
  access.
• To include heuristics, natural language
  processing techniques.
• XSPAN Ontology
• DAMLOIL
• How to provide semi-automatic support for the
  creation of XSPAN ontology?

9
DAI Workshop Show-and-Tell Data Portal (Kerstin
Kleese van Dam, CLRC- Daresbury Laboratory)

• HPC portal to find and manipulate data.
• Wrapper for local database that transforms your
  metadata into a common format.
• Can select data slices using RasDaMan
  multidimensional DB.
• If data requested not in DB, simulation is run
  and data created then archived.
• See http//esc.dl.ac.uk9000/index.html for a
  simple demo.

10
DAI Workshop Show-and-TellDiscovery Net (Vasa
Curcin, Imperial College Inforsense spin off)

• Kensington Data Mining System
• Data mining over heterogeneous data sources
• Bioinformatics pharmaceutical, remote sensing
  data,
• Visualisation methods for interactive discovery.
• Integration into the Grid
• Execute pre-processing and mining procedures in a
  distributed environment.
• Workflow management
• Vasa has offered Geodise a free (time limited?)
  trial of Kensington when we are ready to do data
  mining.

11
DAI Workshop Show-and-TellEuropean Datagrid
(EDG) project Leanne Guy (CERN)

• LHC, raw data reduced through 3 levels from
  40TB/s to 100MB/s.
• Trigger algorithms filter out physics thats not
  important.
• Work package 2 Grid Data Management
• Replication, Grid metadata management service
• Spitfire
• a Relational DB Service for the Grid.
• Decouples the client from the RDBMS backend via a
  mediator.
• API interface with Work package 1 - Computation
• e.g. where is this file? Where is the best place
  to run a job?
• Web services (planned but not yet implemented)

12
UK Role in Open Grid Architecture
Towards an Architectural Road Map (1) Malcolm
Atkinson

• What is OGSA? Web Services Grid technology.

13
UK Role in Open Grid Architecture (OGSA)Towards
an Architectural Road Map (2)

• OGSA describes many things that must be done, but
  not how to do them (1/2 the story).
• OGSA features
• WSDL WSIL WSEL - Description, Discovery
• Tools and Platforms Apache axis,
• Invocation - SOAP, PRC/RMI, Optimised Binding
• Representations XML and XML Schema
• Life time management factories, GS handles
• Authentication, Change management, etc.

14
UK Role in Open Grid Architecture (OGSA)Towards
an Architectural Road Map (3)

• OGSA Development
• Globus2 ? Globus3, end of Dec. 2002
• Need standard model for web service, to achieve
  interoperability.
• Discussion Geodise Condor web service used in
  an example
• Other projects may wrap other job submission
  tools as web services.
• It would be good to have a standard interface to
  job submission services.
• Allows users to shop around.

15
Database Access and Integration Services Planned
Development Activities (Norman Paton)

• OGSA-DAI Project
• Reference implementations docs (first 6 months)
• Core relational services (Brain Collins, IBM)
• Core XML services (Rob Baxter, EPCC)
• Early adopters - AstroGrid and MyGrid at present.
• Questionnaire to be sent round to see which
  projects want to participate/ use software etc.
• Less well defined deliverables over following 12
  months
• Enhancement, specialist data types, distributed
  queries

16
OGSA-Data Access and Integration Services (DAI)

• To add data access integration service to OGSA
• Protocol for accessing DB over OGSA
• Relational/SQL, XML, OO, other semi-structured
  data.
• Transparent integration of multiple databases.
• SOAP/HTTP binding for WSDL protocol.
• Documents
• OGSA, NeSC, DBTF, Globus, GGF, etc
• NeSC, Baseline Grid Data Service Spec Atkinson
  et al, Feb 2002
• DBTF, Paton et al, Watson 2002
• Globus, Tuecke et al, Feb 2002

17
OGSA-DAI (RDBMS) - Brain Collins, IBM Hursley

• Research/Information Gathering
• Documents (see previous slide)
• Globus Toolkit V2 3 GSI, CAS, GridFTP
• Technology
• Linux, SQL, XSQL, JDBC, ODBC, Web Services, Grid
  Java CoG Kit, SRB, etc.
• Scope
• Test with Oracle, DB2, MySQL, welcome feedback on
  other DB.
• Languages (Java/C ) and API (JDBC/ODBC)
• Handle large dataset, multiple protocols for data
  transport.
• Architecture
• DB web service architecture
• 1 DB or multiple DBs (virtual DB) wrapped as web
  services.
• 1st six months concentrate on query and metadata.

18
OGSA-DAI (XML) Rob Baxter, EPCC-Edinburgh (1)

• Grid XML Data Service (GXDS)
• OGSA service wrappers for XML databases
• Plus factory services for management of GXDS
• XML Database Services (currently defines)
• WSDL spec for GXDS version alpha
• query, update
• not implemented yet bulkLoad, schemaUpdate
• Default implementation in Java
• Apache/Axis in Tomcat for hosting environment.
• Apache Xindice (zeen-dee-chay) XML Database
  (supports XPath)
• Possibility of MS .NET in future?

19
OGSA-DAI (XML) - Rob Baxter, EPCC-Edinburgh (2)

• Third party services for data transfer
• Service requesters can use DB services as proxies
• For long-lived queries and large data transfer
• Formal review Aug Sep 2002 (post GGF5)
• Seek feedback from AstroGrid, MyGrid, GridPP,
  RealityGrid, eDIKT
• Geodise to contact Rob Baxter early May for an
  unsupported copy, or at least the WSDLs.

20
DAI Workshop (Day 2, morning)

• Discussion Session
• Security/Access Control Group
• Query Language Group
• Metadata Management Group
• Replication Group
• All groups mainly listed problems/requirements,
  but no solutions offered.
• Future DBTF GGF WG Activities
• Several special interest groups (SIGs) will be
  established, based on the above four discussion
  groups.
• Share ideas between projects.
• Web site/ bulletin board/ mailing list (NeSC to
  create early May)

21

• Security and Access Control Breakout
• Access rights by citizenship in a particular
  group
• May change quickly and be valid for a certain
  period of time.
• Data protection, granting access to your data.
• Are solutions to the problems at a single (RDBMS)
  level, but harder at a federated level.
• Query Languages Breakout
• ADTs how middleware might expose these through
  extensions to standard SQL-92 systems

Query interface
Query in problem domain
Middleware
Publish query capabilities
DBMS
DBMS
DBMS
Translated queries
22
Metadata Management Breakout

• Challenge
• Manage integrity and applicability of mapping(s)
  over time between logical and physical
  representations of the data, and between logical
  and logical.

Superset of definitions within a science
discipline/company
Common reference ontology ?
rules
Semantic structure
Ontology, relationships, meaning of data values.
rules
How to access data without knowing physical
structure. Descriptive name e.g. engineering
optimisation data.
Terminology
1-1 mapping
Internal structure in storage resource. E.g.
ENG_OPT_DAT. E.g. relational tables, XML Schema.
Physical description