1. 17 Dec 2015 Bryan Sullivan, AT&T
Open Source and Info Models
17 Dec 2015
Bryan Sullivan, AT&T

2. Background
OPNFV integrates upstream projects into a coherent platform for NFV, starting from the infrastructure layer and moving up the management stack
OPNFV does not have an information model nor plans to establish one
Relies on existing running code and data models from OpenStack, ODL, etc
OPNFV relies upon upstream open source projects / components which
Expose/consume data via open APIs and specific adapters/translators => flexible
Have implicit models, but not maintained in formal modeling language
Are intended to have limited scope, yet sometimes compete, and evolve toward cleaner boundaries and model convergence over several releases
Jan 2016
ETSI NFV Inter-SDO Information Modeling Workshop

3. Model-Driven Frameworks in Standards and Open Source
Model-driven frameworks are preferred as an architectural approach
Standards based information models are used by open source projects when they meet needs for model-driven management frameworks, e.g. are focused and well-defined
Various model representations will likely have a role for specific areas
YANG: the model representation for netconf-based network config mgmt in ODL
TOSCA: a model representation for cloud services, being applied to NFV
In various cases these may require to/from JSON or YAML for use in specific interfaces / REST APIs
But…
Developers find the path of least resistance and will reuse anything that makes sense to minimize reinventing the wheel
Jan 2016
ETSI NFV Inter-SDO Information Modeling Workshop

4. Example: Lineage of standards adoption for TOSCA
One example of how open source projects pick up draft specs, implement them and (can) provide feedback implementation experience to SDOs, as specs are developed or afterward
OASIS TOSCA
- Large participation from cloud players
- Focused on describing cloud services
2014 – ETSI NFV ISG MANO
Non Normative specification for management and orchestration for NFV
TOSCA For NFV Profile
Evolving
Based on the MANO VNF Descriptor
2015 – NFV IFA
Several normative interface specs re Management and Orchestration, and NFV Information Modeling
TACKER
(Generic VNF Manager for OpenStack)
Note: example only; OPNFV has not yet selected/integrated VNFM solutions
Implementation feedback to published and in-progress specs (a goal…)
Jan 2016
ETSI NFV Inter-SDO Information Modeling Workshop

5. OPNFV Information Model Needs
Rapidly developed, cross-SDO aligned data models
OPNFV is bottom-up, implementation-focused, builds on what is available and fills gaps as needed. R1 (Arno) focused on VIM/NFVI, but OPNFV is expanding to VIM-NBI-using functions e.g. the conceptual VNFM and NFVO. Upstream projects are moving quickly per draft specs for those functions, e.g. VNFM and TOSCA. SDOs should expect gaps to be filled quickly. Thus quick and cross-SDOs aligned progress is required to help ensure that the first generation of NFV platforms leverage standardized information models.
Ability to map NFV concepts, entities, relationships to existing APIs and models in upstream projects
Defacto upstream models may need some “reverse engineering”, e.g. to enable alignment with SDO info models
Focus first on functions of lower layer management stack entities, e.g. VIM and VNFM, then work up the stack and across the VNF/service lifecycle
Open source collaborative energy flows from the bottom up, and for NFV is currently centered around provisioning and management of applications and their resources
Early access and input into the development of industry specifications
As far as possible, all SDOs should make drafts publicly accessible
SDOs and open source projects e.g. OPNFV should share info and feedback through collaboration tools, e.g. JIRA
Jan 2016
ETSI NFV Inter-SDO Information Modeling Workshop

6. Conclusions
CATALYST - OPNFV can be the catalyst for industry adoption of standards based specifications and models if and where they are aligned with the open source implementations
REALITY - SDOs need to account for “de facto” standards both in architecture and in information modeling
TIME - Timely publishing and broad accessibility facilitates alignment – We can’t wait years for a spec to be completed only to realize that it is off from the implementations.
MODULARITY – Publish modules that enable projects to get started soon, without dependency upon a large information model
Jan 2016
ETSI NFV Inter-SDO Information Modeling Workshop

7. We look forward to collaborating with you!
Services motivate development of platforms and standards
We look forward to collaborating with you!
Services
Motivation
Realization
Customer
Choice
Open Source
Platforms
Standards
Platforms are how services are realized and standards validated
Standards provide choice and scalability in platforms and services
Jan 2016