UK e-Science All-Hands Meeting

1
Enterprise specification of theNERC DataGrid

• Andrew Woolf, A.Woolf_at_rl.ac.uk1
• Ray Cramer2, Marta Gutierrez3, Kerstin Kleese van
  Dam1, Siva Kondapalli2, Susan Latham3, Bryan
  Lawrence3, Roy Lowry2, Kevin ONeill1
• 1CCLRC e-Science Centre
• 2British Oceanographic Data Centre
• 3British Amospheric Data Centre

2
Principal partners

• British Atmospheric Data Centre
• British Oceanographic Data Centre
• CCLRC e-Science Centre
• Program for Climate Model Diagnosis and
  Intercomparison (LLNL), EarthSystemGrid

3
NERC DataGrid
British Atmospheric Data Centre
British Oceanographic Data Centre
4
Outline

• Grid identity crisis
• RM-ODP
• NDG Enterprise specification
• Process
• Summary / conclusions

5
The Grid identity crisis

• Need to identify defining patterns of Grid
  computing
• Formal architectural methodologies may help in
  abstracting these patterns
• Requirements capture and analysis is starting
  point of any formal architecture

6
Reference Model for Open Distributed Processing
(RM-ODP) (1)

• ISO 10746-1,2,3,4
• Formal architecture methodology for distributed
  systems
• Viewpoints approach

7
Reference Model for Open Distributed Processing
(RM-ODP) (2)

• Enterprise viewpoint
• roles, activities, policies (incl. VO)
• Information viewpoint
• semantics of information and information
  processing (static, invariant, dynamic schema)
• Computational viewpoint
• interfaces and computational objects (cf. CORBA
  IDL, WSDL portTypes)
• Engineering viewpoint
• distribution infrastructure (e.g. web services,
  WSRF vs OGSI)
• Technology viewpoint
• choices of technology (e.g. app servers, DBMS)

8
Reference Model for Open Distributed Processing
(RM-ODP) (3)

• Enterprise viewpoint
• Specification of purpose, scope and policies of
  system in terms of
• roles identifier for a behaviour
• activities ordered sequence of actions
• policies set of obligations, permissions or
  prohibitions
• Stakeholders must agree on these!
• Structured basis for requirements capture and
  analysis
• UML Use Cases, Collaborations, Activity Diagrams
  may be useful (NB ISO/IEC 19793 Information
  technology Open distributed processing Use of
  UML for ODP system specifications developing UML
  profile)

9
NDG Enterprise specification
10
NDG Enterprise specification Process (1)

• Original plan one major face-to-face meeting for
  each RM-ODP viewpoint
• TOO OPTIMISTIC!
• Finally series of meetings on Enterprise
  specification, i.e. precisely what NDG was
  intended to do
• Requirements refined into EV roles, activities,
  policies between meetings
• Conflicts, assumptions identified for followup
  discussion

11
NDG Enterprise specification Process (2)

• e.g. Security
• clear that role-based system was required
• took considerable discussion amongst all partners
  to settle on details of authorisation model (see
  Lawrence et. al., this meeting)
• e.g. Data discovery and access
• refined into explicit activities
• search ? detailed metadata browse ? data
  browse/selection ? deliver data
• e.g. User Workspace
• identification of EV Workspace Provider role
  raised resource-sharing issues

12
NDG Enterprise specification Process (3)

• Needed involvement of
• resource owners, users, implementers, project
  collaborators
• Not trivial to reach agreement on VO contracts!
• Engagement with experts on current systems
• complexity of real-world datasets
• legacy operational systems
• deployment practicalities, etc...

13
Summary and conclusions (1)

• RM-ODP useful methodology for architecting Grids
• EV provides structured approach to requirements
  capture and analysis (roles, activities,
  policies)
• Need involvement of all stakeholders
• Structured approach helps minimise system scoping
  from drifting to implementation details

14
Summary and conclusions (2)

• RM-ODP viewpoints assist detailed design, e.g.
• NDG EV activities divide naturally along
  metadata/data lines ? core components of
  Information Viewpoint
• NDG Computational Viewpoint interfaces strongly
  motivated by EV activities
• Formal architecture/requirements capture for
  abstracting Grid patterns?