1. Leveraging PIDs for object management in data infrastructures RDA UK Node Workshop, July 16 2019
Tobias Weigel (DKRZ)
- @resdatall
CC BY-SA 4.0

2. https://rd-alliance.org/ - https://twitter.com/resdatall
Motivational use case
CMIP6: Coupled Model Intercomparison Project, phase 6
Data infrastructure support: Earth System Grid Federation (ESGF)
Automated assignment of Handle PIDs as part of e-infrastructure workflow
Specific feature requested by users to improve
tracking of data versions and replicas
referenceability before formal publication
ESGF nodes
Balaji, V., Taylor, K., Juckes, M. et al. (2018): Requirements for a global data infrastructure in support of CMIP6. Geosci. Model Dev. doi: /gmd -

3. Motivational use case: basic workflow
HPC QC
early sharing, community review
B2FIND
PID pages, tools -

4. FAIR Digital Object Management
CRUD
operations
What shall we do if number, variety and complexity of objects increase?
Make tools and services more efficient and effective – create the Intelligent Data Fabric
Automated management of data objects and their metadata
early-to-mid data life cycle – pre-publication
Record connections between (meta)data objects, software, workflows, ... -

5. Combining multiple RDA groups and recommendations
Data Fabric IG:
Architectural vision for DO management
PIDs as key element: e-infrastructure layer of stable references
PID Kernel Information WG:
Required metadata to enable machine actions
Based on work of earlier PID Information Types WG
Data Type Registries WG / Data Typing WG:
Register (meta)data types and provide machine-oriented API
Enable service chaining/orchestration
Research Data Collections WG:
CRUD operations for managing groups of objects -

6. https://rd-alliance.org/ - https://twitter.com/resdatall
Data Fabric IG
Group has existed formally since P4
Concerned with cross-group/discipline infrastructural challenges
Co-chairs: LI Jianhui (CNIC, CAS), Robert Quick (IU), Tobias Weigel (DKRZ)
Aims to promote the creation of an ecosystem of reusable components and approaches built with them
Recognizes that challenges are driven from community demand, but overarching solutions provide larger benefits
Based on an understanding of Digital Objects and PIDs as foundational technologies -

7. RDA Recommendation on PID Kernel Information
7 Guiding Principles for PID Kernel Information
Independent of specific infrastructure or technologies
Geared towards minimizing human interaction, long-term stability of processes relying on Kernel Information
Draft Kernel Information profile
15 elements, 6 aligned with W3C PROV
Exemplary high-level architecture
Use cases and community adoption
x, y, z
Principle 1: The primary purpose of a PID KI record is to serve machine actionable services .
Principle 2: The PID KI record is a non-authoritative source for arbitrary metadata . If the
information for an attribute duplicates metadata maintained elsewhere, the external
source is the authority.
Principle 3: PID Kernel Information is stored directly at the resolving service and not
referenced.
Principle 4: A PID KI record can be changed only by the data object owner or owner delegate
(e.g., PID record manager).
Principle 5: PID KI record values should change infrequently with update initiated only by an
appropriate authority, avoiding human interaction on updates where possible .
Principle 6: Attributes (items) in the profile are expressed as key-value pairs where the values
are simple (indivisible) (first normal form).
Principle 7: Any profile should follow the second and third normal form. Doing so may reduce
migration issues if records need to be migrated to a revised profile in the future. -

8. Unified management of Research Data Collections The RDA Research Data Collections WG Recommendations
Not just describe collections, but enable actions on them
Create, Read, Update, Delete, List plus some others
Machine agents as primary users: built-in scalability and automation
Flexible hierarchy of Services, Collections, and Members
Manage a collection of objects transparently just as another object
API specification against which tools and services can be built across community boundaries
RESTful API specification and elemental metadata model
Pilot implementations in several application areas exist
Collection storage agnostic: Able to work with multiple backends
Multiple points for custom extensions, such as ontologies, typing and specialized operations
(Research) data management beyond single objects
Actions
Create
Read
Update
Delete
List
Service
Collection
Member
Further information:

9. Upcoming: PID Kernel Information Profiles WG
Basic metadata profile defined – community profiles encouraged
Proliferation of profiles is challenge for long-term adoption
How do we manage profiles?
What are good life cycle models for profiles?
Are additional techncial interfaces necessary?
Follow via the group: -

10. Thank you for your attention.-