Page tree
Skip to end of metadata
Go to start of metadata

Models Project Team Meetings

Next Meeting Agenda

2017-Mar-13 Models Project Meeting #28

Agenda:

Minutes of Previous Meetings

2017-Mar-6 Models Project Meeting #27

Attendees:
  • Bryan Sullivan, AT&T
  • Dan Druta, AT&T

  • Aimee Ukasick, AT&T

  • Sridhar Rao
  • Trinath Somanchi, NXP
  • Sridhar Pothuganti, NXP
Agenda/minutes:
  • Danube testing status & stable branch
    • For Danube (stable/newton) verified vHello_Tacker & vHello_3Node_Tacker on Apex (virtual deploy), Packstack (RDO) (bare metal / all-in-one), Devstack (VM)
    • Attempting to find a JOID env to test in
    • Cut stable branch this week
    • Not sure docs will be ready. But no intent to be part of the release.
  • Euphrates release planning
    • Consider overall model-driven NFV goal in OPNFV
      • Policy, Telemetry, Topology/Resources, Application 
    • Add other VNFMs to test
      • Aria
        • Is Aria a VNFM/NFVO or a toolkit to build them
        • If a toolkit, what is the project based upon it that we would actually use
      • Cloudify
      • OpenBaton
      • OSM
      • ONAP / Open-O
    • Evolve toward testing project vs platform component-oriented project
      • Include basic tests and expand to test attribute permutations, e.g.
        • Realtime image import (artifacts:) vs pre-provisioned image (image:)
      • Include in Functest for scenarios with VNFMs
      • Major focus on Hackfests/Plugfests
      • Expand VNF test subjects, e.g. per
        • Prakash will add rows to the VNF use case table at Testing
        • Sample VNF project proposed by Intel
        • VNF Catalog project by LF intern
  • Events
    • Hackfest planning
      • Plan to attend needs justification by specific goals 
      • Charter for Models testing in Plugfests will be strengthened by Dovetail adoption of Models tests
      • Key enablers of hackfest
        • scope of VNFM support across the Models tests
        • scope of reference VNFs we can test
      • We can make a call for adhoc testing for the Paris hackfest but need to make a final decision asap
    • OPNFV Summit planning
      • Any Models-related talks planned for Beijing
      • Models project member attendance is TBD
    • Multi-SDO workshop
      • Planning is TBD
  • Call timing
    • 4PM OK?
      • Call at 00:00 UTC - have not been having it
    • Time zones to consider
      • PST (UTC-8)
      • IST (UTC+4.5)
      • Europe
    • Maybe 6AM PST
      • Merge Domino and Models calls?
      • Bi-weekly alternating?

2017-Feb-27 Models Project Meeting #26

Attendees:
  • Bryan Sullivan, AT&T
  • Dan Druta, AT&T
  • Amir Levy, Gigaspaces
  • Aimee Ukasick, AT&T
Agenda:
  • stable/danube branch plan
    • Plan to cut the stable/danube branch as soon as current baseline tests are updated (this week), with further work on master
  • Hackfest planning: VNF demos/portability
    • VNF portability testing as we had as a goal for the Colorado hackfest (but which didn't happen apart from the Models project demos)
    • OPNFV Danube-based
    • What reference VNFs will be available?
      • updated Models tests: vHello_Tacker, vHello_3Node, vLamp, vHello_Cloudify (updated to Aria if possible)
        • adding new/simple VNFs (if possible)
      • updated vHello_VES
      • Open-O ?
      • OSM ?
      • OpenBaton ?
  • Planning for next Multi-SDO IM workshop has started
  • Testing of modeled VNF capabilities from ONAP - maybe in E-release
  • Gigaspaces is working on Intel's EZ-VNF program on onboarding tools for VNF developers (OASIS WD-04)

2017-Feb-13 Models Project Meeting #25

Attendees:
  • Bryan Sullivan, AT&T
  • Dan Druta, AT&T
  • Krzysztof Frujacz, Gigaspaces
  • Prakash Ramchandran, Huawei
  • Stefano Salsano, Univ of Rome Tor Vergata, working on EU project Superfluidity on NFV
  • Aimee Ukasick, AT&T
Agenda:
  • Test design approach
    • Python vs bash: does it matter? we should use whatever the contributor is most able to contribute. future convergence is an option but the priority is to promote contribution. the goal is upstream contribution of the tests thus if anything the tests should use the framework (and programming env) that the target upstream community uses.
  • Architecture: re the diagram OPNFV Architecture for Danube release
    • It's unclear what the goal of this diagram is, and what the value to OPNFV is of representing its work/product as an "architecture"
    • It's unlikely that many OPNFV projects will be mapped (or can be) to specific components in the diagram
    • It is desirable to know where projects fit, but the mapping is likely to be imperfect
    • probably more useful to consider OPNFV's arch independent of the ETSI arch, e.g. focused on what OPNFV has released in each release, vs a mapping to ETSI (if possible this can be done after the OPNFV arch view is developed)
    • it's also perhaps more useful to focus architecture considerations on the upstream projects and how they are interconnected as architectural components
    • the architecture of OPFNV-driven reference platforms will also evolve over time, e.g. the transition to cloud native may radically change the "architectural view"
    • (Prakash comments on the arch page) In Models meeting discussions it was highlighted that mapping of OPNFV projects to release architecture is more important to the title of the page. Plus the granularity of this diagram is useful but info graphics need not be high priority. Mapping between this and MANO terminology can be taken up later. More important is the upstream architecture and how they get integrated through specific projects in them. Visuals are only cue to perspective and can be many.
  • Status of current work
    • Some updates to the vHello_Tacker test
    • Beyond getting Danube test env running (still working on that), the goals for Danube are just to update the existing tests for Newton
  • Upcoming/Planned
    • Recap of YANG discussion with Peter
      • Requested by Alex

2017-Feb-6 Models Project Meeting #24

Agenda:
  • Status of current work
  • Upcoming/Planned
    • Recap of YANG discussion with Peter
      • Requested by Alex
Attendees:
  • Bryan Sullivan, AT&T
  • Alex Vul, Intel
  • Arthur Berezin, Gigaspaces
  • Amir Levy, Gigaspaces
  • Krzysztof Frujacz, Gigaspaces
  • Trinath Somanchi, NXP
  • Sridhar Pothuganti, NXP
  • Aimee Ukasick, AT&T

Minutes:

  • Status of current work
    • Difficulty with standing up OPNFV Danube test environments has pushed Models out of the Danube release
    • Bryan: trying to get Packstack running with Tacker on stable/newton as a backup, to keep development moving
    • Amir: Aria should be near-ready for testing
  • Taking on discussions for blueprint proposals
    • Alex to provide intro to Open-O data models and Aria use in it
  • Documentation baseline (skeleton) has been setup by Aimee and merged

2017-Jan-23 Models Project Meeting #23

Attendees:
  • Bryan Sullivan, AT&T
  • Amir Levy, Gigaspaces
  • Krzysztof Frujacz, Gigaspaces
  • Trinath Somanchi, NXP
  • Sridhar Pothuganti, NXP
  • Aimee Ukasick, AT&T
  • Chris Lauwers

Minutes:

  • Status of current work
    • vHello_tacker tests on Devstack
      • We are testing against stable/newton.
      • Trinath is trying vHello_Tacker.sh and needs help with setup, e.g. local.conf for Devstack, workstation config.
        • i7, 8-core, 16GB RAM, 1TB disk
        • Xenial VM, 8GB RAM, 20GB disk, 2 cores
    • vHello_tacker tests on OPNFV
  • Upcoming events
    • OpenStack PTG Atlanta
      • Tacker will not meet. Any other models-related meetups needed?
        • Suggest a post to the list if so.
    • ONS
      • ISubmitted by Bryan: Model-Driven NFV in OPNFV Danube

        A key goal for the NFV transformation is that everything possible should be automated through model-based abstractions that enable orchestration and management systems to seamlessly adapt VNFs to a diversity of infrastructure environments. These model-driven NFV goals go well beyond basic resource topologies, e.g. into lifecycle management, various types of policies, and telemetry. In this talk we will cover and demo the OPNFV Danube release support for modeled VNF deployment and lifecycle management through TOSCA-based blueprints, and the roadmap for additional model features.

    • OpenStack Boston
  • Upcoming/Planned
    • Recap of YANG discussion with Peter Lee
      • Requested by Alex

2017-Jan-16 Models Project Meeting #22

Attendees:
  • Bryan Sullivan, AT&T
  • Aimee Ukasick, AT&T
  • Amir Levy, Gigaspaces
  • Arthur Berezin, GigaSpaces
  • Christopher Price, Ericsson
  • Alok Gupta, AT&T
  • Krzysztof Frujacz, Gigaspaces
  • Larry Lamers, VMWare
  • Pasi Vaananen, Stratus
  • Trinath Somanchi, NXP

Minutes:

  • Cloudify/Aria discussion
    • Plan to support all Cloudify plugins
      • Execution plugin is being worked on first (SSH and execution of scripts as "implementation" hooks)
    • How does it manage stateful as well as static blueprint data
      • Both need to be auditable (give me the current state in some easily usable/comparable form, e.g. JSON)
      • Stateful data needs to be reliably stored - is this a database sync issue only?
    • How does it support update of the deployed instances, e.g. to add a new node in the graph or update a node
    • Plans beyond CLI and Python libraries
      • The vision discussed at the ODL Summit was for a RESTful API thru which we can onboard blueprints, deploy, and interact with the deployed instances as objects (e.g. JSON-based), for a completely modeled/object-based abstraction layer
    • Basic VDU properties supported?
    • Discussions with Tacker on reuse?
      • Some time ago, need to refresh the discussion
    • Bryan's goal for Danube is to create an experimental test using Aria as a VNFM for the basic hello-world level test ala vHello_Tacker.sh and the related blueprint tosca-vnfd-hello-world-tacker
      • As soon as there is a stable Aria version that can be used to develop such a starter test, we will work on it.

2017-Jan-09 Models Project Meeting #21

Attendees:
  • Bryan Sullivan, AT&T
  • Amir Levy, Gigaspaces
  • Arthur Berezin, GigaSpaces
  • Christopher Price, Ericsson
  • Krzysztof Frujacz, Gigaspaces
  • Trinath Somanchi, NXP
  • Uri Elzur, Intel
  • Alex Vul, Intel
Minutes:
  • Status of current work
  • Upcoming events
    • OpenStack PTG Atlanta
      • Bryan will be there for Congress and Tacker
    • ONS
      • Bryan intends to submit for Models, VES, and Congress; collaborators welcome
      • OpenCORD / OpenECOMP will also factor
    • OpenStack Boston
      • Similar plans, different spin
  • Upcoming/Planned
    • Cloudify/ARIA discussion
      • Next week; effort to create building blocks for orchestrators, as a new library compatible with TOSCA
    • Recap of YANG discussion with Peter
      • Requested by Alex

2016-Nov-28 Models Project Meeting #20

Attendees:
  • Bryan Sullivan, AT&T
  • Trinath Somanchi, NXP
  • Sridhar Pothuganti, NXP
  • Stefano Salsano, Univ of Rome Tor Vergata, working on EU project Superfluidity on NFV
  • Giuseppe Carella, Fraunhofer Fokus
  • Dan Druta, AT&T
  • Arthur Berezin, GigaSpaces
  • Alex Vul, Intel
Agenda:
  • Status of current work
    • Updates to Comparison page
      • (Dan) TOSCA Simple Profile is a realization of the ETSI Information Model, based upon an earlier version. Question is whether they (question) need to be compatible, or complementary.
      • TOSCA version of ETSI model is in development (by who?)
      • (Alex) OASIS's TOSCA meta-model is more generic.
      • (Arthur): ETSI model for VNFs and OASIS's Simple Profile are ~2 years (Arthur: Correction 10 months)old and need to be aligned with the new ETSI Information Model.
      • (Alex) recommend that OSM be brought in to the discussion, e.g. Adrian (his input will be provided in Bonn at the Multi-SDO Modeling workshop).
      • (Giuseppe): we should also connect with the planners of the ETSI Plugtest in Jan. Vendors are providing descriptors for the different MANO implementations.
      • (Stefano): suggest to add an ETSI column. Stefano will edit it.
      • (Dan): a profile of the ETSI model will be needed to apply the Info Model to a particular domain language e.g. TOSCA.
      • (Arthur) TOSCA is an OO descriptor language. A TOSCA template can be created without a profile, by importing the node types to a model.
      • (Alex): The Simple Profile should not be required to model a VNF.
      • (Dan): Having a consistent representation is important.
      • (Alex): The core model aspects should be reusable across different domains.
      • (Bryan): who do we look to to drive that domain-specific implementation (open source, standards).
      • ETSI is working on the new profile version which will include updated mappings to TOSCA.
    • Test updates
  • Upcoming events
    • Plugfest planning
      • Attendance at hackfest session
  • Upcoming/Planned
    • Cloudify/ARIA discussion
    • Recap of YANG discussion with Peter

2016-Nov-21 Models Project Meeting #19

Attendees:
  • Bryan Sullivan, AT&T
  • Sridhar Pothuganti, NXP
  • Prakash Ramchandran, Huawei
  • Peter Lee, Corenova
  • Dmytryo Gasanov, Netcracker
  • Ulas Kozat, Huawei
Agenda:
  • Recap YANG discussion
    • (Alex) Design-time vs run-time modeling; TOSCA is suited to design-time modeling, but existing models (e.g. YANG) perhaps better for operational state.
      • Design-time aspects: resources, topology, lifecycle events from a infrastructure (service, function) management perspective. Holistic (addresses all needs lifecycle-centric). To start with, the design is generic although specifics may be needed for functions (e.g. image).
      • Run-time aspects: as-deployed/operational state, telemetry, changes to the service/function (resources, topology (e.g. placement), ...). Representation will diverge from the design-time model (while remaining consistent with the design goals) as best benefits the purpose of maintaining the operational state.
    • (Alex) future discussion: MEF, TMF SID (question), Django views of service modeling. What is the right way, e.g. for interoperability?
    • Bryan will invite Peter back for a followup.
    • Bryan will invite Francisco-Javier or Adrian to give an overview of OSM modeling.
  • Getting involved
    • Status of current work
    • Help fill out the Comparison table, with aspects of modeling related to the projects you are involved in.
    • Develop tests
  • Upcoming events
  • Upcoming/Planned
    • Cloudify/ARIA discussion

2016-Nov-14 Models Project Meeting #18

Attendees:
  • Bryan Sullivan, AT&T
  • Aimee Ukasick, AT&T
  • Amir Levy, GigaSpaces
  • Deepak S, Intel
  • Guiseppe Carella, Fraunhofer Fokus
  • Sridhar Pothuganti, NXP
  • Trinath Somanchi, NXP
  • Alex Vul, Intel
Agenda:
  • Status of current work
  • YANG Models discussion with Peter Lee
    • Peter described OpenCORD approach - see the References page for notes on OpenCORD.
  • Modeling approaches - collecting perspectives on
    • What ETSI is recommending in their current drafts and why

      • Bryan will reach out to Francisco Javier for feedback.
    • The role of YANG in service/application control

      • Seems there is significant consensus on this, e.g. in OpenCORD and ECOMP approaches. Details will be added to the wiki.
    • How service and VNF modeling works in OpenCORD today

      • Additional details will be added to the wiki based upon the references Peter provided.
    • If YANG was used at the service and VNF layers, how we can derive TOSCA NSDs/VNFDs etc as needed for processing by the cloud controller components e.g. Tacker?

      • We seem to have a common overall goal of a model-based abstraction that can be interacted with by upper-MANO stack components (or other service components), which is then by some service/application control framework, is bound to specific actions to be taken at infrastructure managers (e.g. cloud, SDNC, or application-directly-exposed APIs). Examples of this approach will be added to the wiki, e.g. ECOMP's ODL-based Application Controller, or OpenCORD.
      • We need to describe (for different projects/approaches) the overall process/toolchains for taking a high-level service/VNF abstraction thru the transformations needed to execute specific infrastructure management actions. That should help us come to a common understanding at least.
  • Planning for Danube
  • Upcoming events
    • Plugfest use cases (reference VNFs)
  • Upcoming/Planned
    • Cloudify/ARIA discussion

2016-Nov-7 Models Project Meeting #17

Attendees:
  • Bryan Sullivan, AT&T
  • Prakash Ramchandran, Huawei
  • Larry Lamers, VMWare
  • Dan Druta, AT&T
  • Deepak S, Intel
  • Sridhar Pothuganti, NXP
  • Steve Vandris, Intel
  • Aimee Ukasick, AT&T
  • Trevor Cooper, Intel
  • Andrew Veitch, Netcracker
  • Alok Gupta, AT&T
Agenda:
  • Status of current work
    • (bryan) Implemented 3node test
    • Near-term goals:
      • Update to Newton (JOID, Apex)
      • Expand beyond Tacker, install MANO projects as a black box
      • Prepare for Plugfest, add some more blueprint attributes
    • Prakash: Open-O supports install in docker, also JuJu (as a VNFM uses 2.0)
    • Resolve open question: can multiple VNFMs co-exist
  • Planning for Danube
    • Should be done as of now
    • Next milestone is "(MS2) Detailed test case descriptions communicated to test project teams"
      • Need to prepare expected scope and test descriptions
  • Upcoming events
    • Plugfest use cases (reference VNFs)
    • 2nd call for participation https://lists.opnfv.org/pipermail/opnfv-tech-discuss/2016-November/013509.html
    • Assume plugfest will proceed wit limited support
      • Open-O should be there (participation in VNF Portability is TBD): Lingli and Chris
    • POD resources: what resources will be needed should be known asap
      • Update the Resource Requirements section of the plugfest page
  • Models Team Meeting B at 0000 UTC
    • Will start the bridge etc
  • Status of committer pool etc
    • Bryan so far has been the sole active contributor/committer to the Models repo. Additional reviewers/committers are welcome to get involved in the project, and I can slow down the speed of commits as needed to enable team members to come on board. More active committers can also help establish more conventional practice for commits etc, a good thing for the project.
  • Next week
    • Cloudify/ARIA discussion
    • YANG Models discussion with Peter Lee

2016-Oct-31 Models Project Meeting #16

Attendees:
  • Bryan Sullivan, AT&T
  • Prakash Ramchandran, Huawei
  • Larry Lamers, VMWare
  • Dan Druta, AT&T
Agenda/Minutes:
  • Dec Plugfest planning
    • Finalize this week:
      • Who can bring VNFs/VNFMs
        • vIMS in Open-O and OSM via JuJu
      • Who is ready to test against Models "Hello World" blueprint 
      • Get confirmations of attendance (in person, virtual) this week
  • Planning for Danube
  • Recent events
    • OpenStack Barcelona demos / hackfest sessions
      • Collecting best links from demos etc on the Models wiki
      • Doctor/Vitrage/Models
      • Domino
      • VES
    • TSC/BoD takeaways
      • "Architecture of OPNFV" re CVP program and goals for what we can claim in 1st release of CVP
        • Where how this will be addressed is TBD
        • MANO WG should be a good venue for it
        • Prakash has a draft diagram; some open issues e.g. underlay arrangement, service assurance
        • Dan: 2 levels i.e. VNF and service; testing discussions should clarify the VNF and service design handoff, and that VNFs are reusable; both VNFs and services can be onboarded. Tacker as an example has this distinction and two separate catalogs for them.
      • MANO WG intro
        • Few comments overall
        • Clarified that the role of Models project is to "assess and advise"
        • Send out a notice for the PM PST timeslot for Models; Prakash will send for MANO WG as well.
    • Charmers meetup
      • Since JuJu is a common MANO stack element across projects, being familiar with its DSL is important
  • Cross-project
    • Movie project status
      • ONOS J release will have YANG model support
      • use cases include Virtual Private Cloud, L3VPN, vEPC, Bandwidth on Demand
      • ODL has NEMO project, being ported to ONOS by Movie project
  • Upcoming
    • Cloudify/ARIA discussion

2016-Oct-17 Models Project Meeting #15

Attendees:
  • Bryan Sullivan, AT&T
  • Guiseppe Carella, Fraunhofer Fokus
  • Prakash Ramchandran, Huawei
  • Michael Bugenhagen, Century Link
  • Larry Lamers, VMWare
  • Amir Levy, Gigaspaces
  • Dan Druta, AT&T
  • Peter Lee, Corenova
  • Alex Vul, Intel
Minutes:
  • Status of current work
    • Bryan showed the current work on the vHello_VES model, with App VM (nginx-based web server), VES Monitor VM. Will add LB VM (iptables), a 2nd App VM. This will enable us to add scaling to the blueprint, through TOSCA scaling policies (hopefully supported). The VES demo will include various lifecycle events for which the intended handling should (if supported) have some modeled aspects, but for now will have scripted handlers at least.
  • Planning for Danube
    • JIRA will be finalized by next Tuesday, comments welcome
  • Cloudify/ARIA discussion
    • Amir presents an ARIA introduction
    • ARIA adds TOSCA node types that are processed by ARIA library functions which interact with VNF/service orchestration
    • Prakash: add Domino as a project that can leverage
    • Mike: industry needs to develop a common platform model at the technology layer, which is mappable to business models
    • Alex: what's missing in TOSCA?
      • Mike: Custom topologies cause special work for each provider (lack of interoperability)
    • Bryan: will add this scope aspect to the wiki
      • alignment with external work e.g. BBF/MEF/3GPP on the common platform model
      • once open source implementations are available, consider demos of how the common model improves interoperability as part of the models scope
    • Mike: Service Catalogs also need a model-driven framework
      • The smallest purchasable item is the atomic service that needs to be modeled
      • MEF is experiencing related issues
    • Roadmap
      • TOSCA <> YANG Model: Michael Brenner is leading work, Cloudify has YANG-TOSCA integration, don't believe one will replace the other.
        • Bryan will add a diagram illustrating how these are related in AT&T, e.g. YANG use in the ODL-based APP-C
      • Execution plan: will allow users and providers to understand and as needed negotiate resources
        • Alex: this can help side-by-side comparison
        • Mike: without a top-down definition, there will be interop issues
        • Bryan: should such a plan be derivable from the blueprint, or an additional artifact/node ?
          • Alex: some companies have done work e.g. in "reverse planning"
          • Bryan: a more direct approach would be to model the execution plan
        • Mike: need to start with a schema/taxonomy for the goals of the execution plan
    • Next steps for ARIA
      • When can we integrate ARIA in a Models use case test?
      • We will have a separate side discussion
  • Input for TSC
    • Role of the Models project in helping flesh out MANO integration into OPNFV
  • Upcoming events
  • Next week

2016-Oct-10 Models Project Meeting #14

Attendees:
  • Bryan Sullivan, AT&T
  • Larry Lamers, VMWare
  • Prakash Ramchandran, Huawei
  • Dan Druta, AT&T
  • Uri Elzur, Intel
  • Alex Vul, Intel
Minutes:
  • Status of current work
    • VNFMs in testing
    • VNFM support planned
  • Planning for Danube
    • Release plan: Release Plan
      • Edit etc as needed, we need to agree by 10/25
      • Add other items here in the minutes if desired, or as JIRA items
    • JIRA issues
    • Package format
    • Clarify what "VNF modeling" means in detail
    • Desired end-state, e.g. nirvana of a model-driven config/state representation that can be manipulated etc
    • Open-O VNFM/NFVO, e.g. ARIA+JuJu
  • Upcoming events
    • OpenStack Barcelona demos / hackfest sessions
      • VES+Lampstack demo at the AT&T booth
      • Models meetup possibility, or discussions at the OPNFV booth (we can co-signup)
      • Domino demo
      • Session recommendations related to modeling and Domino
    • Plugfest use cases (reference VNFs)
      • Bryan will send out invites to VNFM/NFVO projects
      • Agree on strawman reference VNFs, at least a simple case e.g. vHello, extended as prioritzed for essential/high-priority aspects of the blueprint
  • Next week
    • Cloudify/ARIA discussion

2016-Oct-3 Models Project Meeting #13

Attendees:
  • Bryan Sullivan, AT&T
  • Larry Lamers, VMWare
  • Amir Levy, Cloudify
  • Prakash Ramchandran, Huawei
  • Dan Druta, AT&T

Agenda:
  • Status of current work
    • Cloudify Manager vHello is working, Bryan will document the issue that was overcome
      • Bryan will setup a debug session with Cloudify team in NA
    • VES data modeling approach will be further analyzed, and how the model fits into the VNF/service descriptors etc
  • Takeaways from ODL Summit and OpenStack Seattle Days
  • Danube project planning
    • Modeling the injection of parameters into VMs, e.g. for "VES support in blueprints"
    • Bryan will propose some Epics, Stories, Tasks etc
      • The team is encouraged to watch, comment, add more, etc
  • Cross-project items
    • Catalog: the OpenStack Community App Catalog and its Horizon plugin used by Murano etc; for Models and OPNFV more broadly, we should use this feature (integrated into Danube) as a place to reference (and provide users the ability to launch) the VNFs that have been tested on the OPNFV platform
    • Domino project Cross-WAN L3VPN use case: Domino includes some use cases for NSD distribution across multiple sites, with elements (VNFs) of a service distributed across the sites, and thus the need for connection point links setup across those sites, e.g. via L3VPN. Extracting the L3VPN component of that goal and developing it as a Models use case seems useful - for the two cases:
      • Two sites, with VMs interconnected via L3VPN: basic ability to connect two VMs via L3VPN, remotely
      • Single site, with VMs connected via L3VPN: basic ability to connect two VMs via L3VPN, locally
      • Two sites, with VMs connected locally and remotely via L3VPN: this might apply e.g. for VNF redundancy or load balancing across sites
    • Modeling VES data model, e.g. in UML or other formats and related tools

2016-Sep-19 Models Project Meeting #12

Attendees:
  • Bryan Sullivan, AT&T
  • Prakash Ramchandran, Huawei
  • Artur Tyloch, Canonical
  • Ulas Kozat, Huawei
  • Alex Vul, Intel
Agenda:
  • Status of current work
    • Hello World on Tacker
    • Hello World on Cloudify
  • VNF Onboarding
  • VES event model in YANG
  • Goals thru 2016
    • Demo at OpenStack Barcelona
    • Dec plugfest at U-NH
    • Reference libraries
Minutes (feel free to add additional notes):
  • Status of current work (Bryan)
    • Hello World on Tacker: vHello_Tacker.sh
      • Tacker install and blueprint deployment is working reliably on Apex and JOID, in non-HA/no-SDN scenarios.
    • Hello World on Cloudify: vHello_Cloudify.sh
      • Cloudify-CLI test development is suspended due to apparent incompatibility with OpenStack Mitaka (the Cloudify OpenStack Plugin is not compatible with recent OpenStack releases)
      • Cloudify-Manager test development continues. Currently the installer (cloudify-setup.sh) is working but starting the blueprint is failing (timeout installing the Cloudify Agent n the VM... being investigated).
    • VES support in blueprints
      • A quick investigation was made into how a VNF's VDU can discover its Nova instance ID, so that it can attach to the VES Collector and identify itself thru the ID. It was found that two methods exist for this: the Nova boot "meta" option and the Nova config-drive option. The metadata service however does not provide the actual Nova instance ID (some other value is in the "instance id" field). The config-drive option does. See the attached vm-config-demo.docx demo log file (which combines a demo of the Hello World Tacker test and the Copper smoke01.sh test which now supports creation of the config-drive for one of the VMs).
      • The key takeaway for the Models project on this is that it represents one of the things that needs to be model-able, i.e. that in some cases a VNF needs to know its Nova instance ID (or other metadata), and thus some way to indicate that the config-drive should be allocated (or the metadata provided through the metadata service). This need should be expressable in the TOSCA blueprint for a VNF.
      • Alex questioned the approach of providing metadata of this type directly to VMs, and offered some other suggestions on how to achieve similar goals. Those discussions will continue as needed in the VES project, which is where modeled telemetry/closed-loop-monitoring will be addressed.
  • VNF Onboarding (Bryan)
  • VES event model in YANG (Bryan)
    • I am working on a YANG data model for VES, and will upload that to the VES wiki and git repo soon. The telemetry needs/capabilities of VNFs will be a model-able aspect that needs to be considered as part of the VNF package. This is the first case (in OPNFV AFAICT) for a YANG model to be part of a VNF blueprint (others are expected soon e.g. for application data model and lifecycle events). For the Models project, the issue will be how to include YANG data models in the VNF package, and what NFVO/VNFM functions will handle them. 
  • Goals thru 2016 (Bryan)
    • Demo at OpenStack Barcelona
      • As a sponsor, AT&T will have a booth at OpenStack. We will demo a VNF running on OPNFV Colorado, and integrating with a PoC closed-loop-monitoring system using VES as the event stream framework. This demo will include a reference VNF installed via a TOSCA blueprint via Tacker or Cloudify (or both, if we have time).
    • Dec plugfest at U-NH
      • As shown at Colorado Plugfest Test Cases, the December plugfest will cover various reference blueprint tests with any VNFM/NFVO projects/vendors that want to demonstrate their compatibility with those reference blueprints. The blueprints will be provided in October so the VNFM/NFVO under test can determine whether they will be able to participate. As noted, the goal is not 100% compatibility verification, rather "to demonstrate the degree of portability, uncover issues for followup, build the library of tested blueprints (VNFM-specific, as needed), and overall come away with a much clearer assessment of VNFM product support for blueprint standards"
    • Reference libraries
      • These will be developed as part of the Models repo through Q4 2016. The goal is to establish several reference VNFs and blueprints, for use in plugfests and Functest.


2016-July-31 Models Project Meeting #11

Attendees:
  • Bryan Sullivan, AT&T
  • Guiseppe Carella, Fraunhofer Fokus
  • Larry Lamers, VMWare
  • Prakash Ramchandran, Huawei
  • Dan Druta, AT&T
  • Alok Gupta, AT&T
  • Artur Tyloch, Canonical
  • Uri Elzur, Intel
Agenda:
  • Status of Colorado release work
  • Takeaways from the OPNFV Summit
    • Modeling in the broader topic of VNF Onboarding
      • In the end-to-end lifecycle of VNFs/Services, the onboarding phase needs to be broken down 
      • Developer experience
        • Modeling/packaging tools for developers
      • Publication to service provider
      • Validation and testing
      • Augmentation of the models/blueprints by the service provider
      • Cataloging
      • Deployment
      • Guiseppe: OpenBaton has some experience with the process
    • Reference VNF library to help drive model consistency/interoperability
  • Role of Models project in the MANO WG
    • Models will collect MANO project info and drive FuncTest related to the common models that we collect and prepare
    • Models will provide a funnel for feedback to SDOs based upon the convergence discussions
    • Models will address various goals for models use, e.g. deployment, application lifecycle, and operations/analytics
  • Guiseppe: preparing a project proposal for OpenBaton
  • Narinder mentioned on IRC: JuJu has been updated to run as a VNFM under OpenStack, similar to Cloudify
  • Should Models project meeting merged with MANO WG
    • The team agrees that this would be a good thing
    • The current time is under discussion; maybe use the Models timeslot every other week
    • Prakash

2016-July-25 Models Project Meeting #10

Attendees:
  • Bryan Sullivan, AT&T
  • Guiseppe Carella, Fraunhofer Fokus
  • Larry Lamers, VMWare
  • Marc Flauw, HPE
  • Prakash Ramchandran, Huawei
  • Alok Gupta, AT&T
  • Artur Tyloch, Canonical
  • Narinder Gupta, Canonical
Agenda:
  • Status of Colorado release work
    • Bryan is making progress on the vHello test case; currently testing Cloudify Manager install
    • Other team members could extend this to other VNFMs as time permits
    • Bryan will create a table on the wiki to show who is investigating different VNFMs, and assumptions/approaches to deploying the VNFM in Functest
    • We can create variations of the vHello test case using different VNFMs
    • Artur: Open-O approach is to use TOSCA for the service orchestrator (Cloudify) and JuJu is the VNFM, not using TOSCA
      • VNF deployment and management probably will be sequentially addressed in the Models project
      • We can test permutations of these also
  • Takeaways from the OPNFV Summit
    • Modeling in the broader topic of VNF Onboarding
    • Modeling/packaging tools for developers
    • Reference VNF library to help drive model consistency/interoperability
  • Role of Models project in the MANO WG
2016-June-6 Models Project Meeting #9
Attendees:
  • Bryan Sullivan, AT&T
  • Artur Tyloch, Canonical
  • Narinder Gupta, Canonical
  • Marc Flauw, HPE
Agenda:
  • OPNFV Berlin Summit
    • The two session proposals by Uri and Bryan are merged into one, titled: MANO modeling: Service, Infrastructure and VNF On-boarding and Interoperability

      •  “MANO is one of the critical layers for NFV realization, yet industry direction on it is in danger of diverging. Along with closed implementations, open source projects such as OSM and Open-O are now driving MANO implementation. To help drive broad convergence and ease Service development, VNF on-boarding, and interoperability, we need broad agreement on Information Models and Data models. We will present a proposal to use TOSCA/YANG as a basis for this, and report on efforts in OPNFV to validate and promote convergence in support for model-driven Service and VNF deployment on the OPNFV platform. This talk will be supplemented by a demo of the current work in OPNFV to be presented in the demo theater.

      • Bryan will upload a draft of the second part of this talk this week for review.
  • Focus on the data models that developers will be expected to prepare for onboarding.

    • Existing vIMS assumes a VNFD element within a overall TOSCA-based blueprint
    • This may be translated into another data model for use by elements e.g. VNFM
  • Focus on the process of ingestion e.g. any assumptions on translation
    • what projects will do the translation, e.g. ARIA
    • e.g. sync between VNFM and NFVO on running state, and any needs for interfaces or backward translation supporting this
  • Focus on service modeling
    • orchestration of resources across distributed clouds and WAN
    • application lifecycle aspects, e.g. scripts triggered by lifecycle events vs modeled e.g. in YANG
  • Relation to OPNFV MANO project proposal
    • TBD; perhaps Models can help setup some basic scenario tests of use to the project
  • Test examples for Berlin
  • Describe some of the implications of the different modeling approaches and tools

2016-May-23 Models Project Meeting #8

Attendees:
  • Bryan Sullivan, AT&T
  • Artur Tyloch, Canonical
  • Marc Flauw, HPE
Agenda:
  • Intro to Models project for Artur (Bryan)
    • First goal is to have a working simplified blueprint that we can drive with multiple VNFMs/NFVOs in functest
    • Real VNFs would be great but are optional.
  • Status of OSM (Artur)
    • OSM Release 1 (see wiki) will be released over the coming months. Release 0 is what was presented at MWC.
    • The role of components in OSM (principally rift.ware and Juju) will be evolving, with Juju taking on most of the VNFM functionality including YANG-based application control/lifecycle (something not yet considered in ETSI, but important overall for NFV).
  • Status of ETSI specs (Marc)
    • Still planning stage 3, current stage 2 specs go into more detail but not yet fleshed out at stage 3 level
    • IFA011 (VNFD) and IFA014 (NSD) will be furthered as stage 3 models in the SOL group
    • IFA007 (NFVO) and IFA008 (EM) (VNFM interfaces) will also be modeled in the SOL
    • IFA005 and IFA006 (VIM interfaces) - plans for modeling are TBD, e.g. delta with OpenStack
    • IFA does not yet consider the role of SDNCs (considered as part of EVE005)
    • Lifecycle management of VNF software is driving consideration of identifiers for VNFDs etc as the software will change
    • IFA015 draft 040 will been upload

2016-May-16 Models Project Meeting #7

Attendees:
  • Bryan Sullivan, AT&T
Agenda:
  • Model analysis status
  • Tools analysis status
  • Model testing status
  • Upcoming events
    • OPNFV Design Summit
    • OPNFV Summit
      • The two session proposals by Uri and Bryan are merged into one, titled: MANO modeling: Service, Infrastructure and VNF On-boarding and Interoperability
        •  “MANO is one of the critical layers for NFV realization, yet industry direction on it is in danger of diverging. Along with closed implementations, open source projects such as OSM and Open-O are now driving MANO implementation. To help drive broad convergence and ease Service development, VNF on-boarding, and interoperability, we need broad agreement on Information Models and Data models. We will present a proposal to use TOSCA/YANG as a basis for this, and report on efforts in OPNFV to validate and promote convergence in support for model-driven Service and VNF deployment on the OPNFV platform. This talk will be supplemented by a demo of the current work in OPNFV to be presented in the demo theater.”


2016-May-9 Models Project Meeting #7

Attendees:
  • Bryan Sullivan, AT&T
  • Mark Flauw, HPE
Agenda:
  • Review of previous meeting
  • Updates to epics/stories in Jira
  • R3 project plan and minimum goals
  • Model analysis status
  • Tools analysis status
  • Model testing status
  • Recent events
    • NFV World Congress
    • OpenStack
  • Upcoming events
    • OPNFV Summit

Minutes (#A) :
Reviewed the last meeting's results since it has been three weeks.
The JIRA epics/stories will be updated with additional info as requested by Gergely, so the nature of the tasks is clearer.
  • But overall the intent in this project is not to over-depend upon the JIRA tool for project tracking, as the nature of the project is different from one that is developing a lot of code.
Plans for Colorado at this point are to:
  1. Generate a test framework for model validation across multiple VNFMs, and include this in FuncTest
  2. High-level focus will be on assessing the state of IM at ETSI and relation to the DSLs of TOSCA, YANG, etc, and include this first in the wiki, and then in a report.

Marc:

  • ETSI specs and models are UML-based (IFA015), IFA011 goes into details such as element hierarchy in a text/table presentation, still at stage 2.
  • The SOL group is codifying the stage 3, e.g. SOL001 (TOSCA-based NFV NSD).
    • NFV-SOL001: Network Functions Virtualisation (NFV); Protocols and Data Models. NFV Descriptors based on TOSCA; TOSCA-based NFV descriptors spec

    • The scope of the SOL work is to develop a data model specification for NFV descriptors fulfilling the requirements specified in GS NFV IFA 011 and NFV GS NFV IFA 014. The specification will be based on the Simple TOSCA profile specification for NFV with possible changes. The deliverable will contain normative provisions and an informative
    • Marc made a presentation to the ETSI Plenary on the IM status: https://docbox.etsi.org/ISG/NFV/05-CONTRIBUTIONS/2016//NFV(16)000168r1_NFV_Information_Model_Status____Relationship_with_other_SDOs.pptx

Bryan:

  • In OMA the Service Enabler Virtualization (Seville) work item is developing a white paper on how OMA specs/enablers can be virtualized thru modeling at the service and VNF scopes.
    • http://member.openmobilealliance.org/ftp/Public_documents/TP/2016/OMA-TP-2016-0061-INP_Seville_WI_Wrapup.zip describes the state of the work, which I am wrapping up due to shift in focus on OPNFV over the last year.
    • The paper and and deck describe how a use case (OMA Secure User Plan Location (SUPL)) might look as a TOSCA-based blueprint. I will be using and further developing this as an example in the Models project.
    • Other OMA enablers to be used as examples include the OMA Device Management enabler Lightweight Machine-to-Machine (LWM2M), which defines a IoT-focused DM framework for constrained devices/networks (e.g. sensors and other industrial/consumer devices over LPWA access). OMA LWM2M is the lead technology being developed in the IoT market for managing and accessing data of IoT devices. 
    • In OpenDaylight there is a project to support IoT DM and OMA LWM2M thru YANG modeling. I submitted a BoF proposal for the ODL Summit on this (LWM2M and ODL): http://events.linuxfoundation.org/cfp/proposals/7061/9879

  • Other than verifying ability to launch an enabler via a modeled topology, these use cases depend upon open source implementations of the actual enabler components.

    • In the meantime I am working with implementers to support the use case testing thru their proprietary implementations, and also looking to use existing open source enabler implementations as noted on the Models wiki.

  • A key question is whether we need to look at these use cases from a DSL view or a higher level view


Marc:

  • Suggest to start at the DSL level for use cases as the UML/IM level is intended to be more conceptual and language independent, even use-case independent (the classes do not represent use case considerations rather generic usable concepts that can be implemented thru the DSL)
  • IM's can help to assess the consistency of the little pieces of the problem space, e.g. classes, e.g. services, service chain, etc...
  • The IM should also enable the relationship description at the UML level and below.
    • e.g. the concepts of client and server in context of a particular service

Bryan:

  • It may thus be useful to start at the UML level for service use cases that may be applied to different service environments, e.g. a web service that can be deployed for browsing, APIs, end-users, and IoT applications
  • Such a web service might include a proxy, cache, and other network functions that can be modeled in terms of their function, relationships, and position in a service topology

2016-April-11 Models Project Meeting #6

Attendees:
  • Bryan Sullivan, AT&T
  • Prakash Ramchandran, Futurewei
  • Gergely Csatari, Nokia
  • Ulas Kozat, Huawei
  • Tianran Zhou, Huawei
Agenda:
  • Selecting a new meeting time
  • Contributors to various epics/stories
  • R3 project plan and minimum goals
  • Events plan
  • Model analysis status
  • Tools analysis status
  • Model testing status

NOTE: Bryan got the meeting time wrong and started an hour too early, thus Marc was not able to join. Models #A Meetings starting next week will be at the expected time (16:00 GMT).

Minutes (#A) :
Prakash requested clarification of what types of model validation (e.g. mulit-node cluster, network links for relationships) will be supported.
  • Bryan: That's a question we need to address thru use case focus. The existing VNF implementations will be used as a template for other use cases, more narrowly focused (my suggestion). 
  • Csatari: The use cases refer to reference VNFs, but should be more abstract e.g. scaling (would like to see the use case description updated)
  • Bryan:  they can be to start, but to drive specific model analysis or test activity, they need to get more specific
  • Prakash: re scaling, I think the best will be take simple VNF scaling with single VDU and another with 2 VDUs and get some models and see how it pans out
Prakash: continuing on the subject lets take Simple TOSCA model for VNF and compare that with YANG or UML, and see what the scaling parameters are. I am looking for scaling attribute in VNF if it is usable from TOSCA or YANG and how do they differ for a VNF scaling use case. See http://docs.oasis-open.org/tosca/tosca-nfv/v1.0/tosca-nfv-v1.0.pdf
  • Ulas: current models are simple, as a scale attribute one can specify number of instances. At the end of the day, these are all {key,value} pairs If you want to define a scaling property in yang, is there a limitation? It is up to you how you define new nodes, policies, properties, etc. If you want to scale a collection of things in Tosca, you can group them and attach a scaling policy
  • Prakash: I think we have to start with use case scaling is just an abstraction , so what is then the use case not scaling but a service and modeling service will inolve first lets say network service consists of one VNF with one VDU and so on. See Policy Example - I; under policies tab "mu_anti_collocation_polic" is something quite arbitrary
  • Ulas: see Policy Definitions, this is actually more informative. For instance you can define a VNF (which is a node) composed of multiple VDUs and then define some properties to represent the bandwidth constraints between VDUs. Then you can specify some triggers based on the values of these properties.
  • Prakash: we need to look at models as what they represent based on Information models and not data models; For a given use case we need to start at Common Information Model, but then that still will need topology representation. So TOSCA puts models in terms of VNF, and now where as YANG puts it in terms of network node, is that a fair statement?
  • Ulas: It is up to the model author to be generic or specific (informational vs. data model) and both Tosca and yang can deliver information and data models. One can also describe the internal workings of a high level service also in yang, isn't this the case?
  • Prakash: Abstract vs. concrete is really the difference so we need to choose an abstract, concrete or hybrid way to model logical network service.
  • Ulas: Currently Tosca is a mixture, e.g. you can specify OS and hardware in Tosca; you can also leave it at "flavor". A VDU can still be abstract, as a unit of scale you need to add/remove resources at that granularity.
  • Prakash: VDU is already mixing the concrete and abstract in TOSCA. A Flavor is good for Compute service in OpenStack, still evolving in Network Services in Nova, so we will need to agree on some basic elements for both TOSCA and YANG as two main modeling for network Services, for Applications it has always been UML.
Minutes (#B)
Bryan described the Models 1st release focus per the last meeting notes (see "minimum goal for R3" below). In summary using the complex VNF model described by https://github.com/Orange-OpenSource/opnfv-cloudify-clearwater/ as a starting example, we will create a simpler one as an extensible template for model testing in the Colorado release, using various VNFMs to assess the degree of model portability.
Tianran intro'd his involvement in IETF etc, for models in MANO area. His team is involved in "Carrier SDN" (generic term) work in TOSCA.
We discussed the Models project relation to the Movie project. There are various details of the Movie project that relate to the Models project and also to Copper (e.g. the Congress blueprint on policy abstraction by Yali).
  • Bryan will review the Movie project details and suggest specific cross-project efforts we can start.
Tianran introduced the NeMo project and the NeMo ODL project. The policy abstraction described there is directly applicable to the Models project, and can be a test use case to pursue.
There is also relation to the ONF's Project Boulder


2016-April-11 Models Project Meeting #5

Attendees:
  • Bryan Sullivan, AT&T
  •  
Agenda:
  • Selecting a new meeting time
  • Status of project setup: Git, JIRA, Wiki
  • JIRA epics and story overview
  • Contributors to various epics/stories
  • R3 project plan and minimum goals
  • Events plan
  • Model analysis status
  • Tools analysis status
  • Model testing status

(Bryan's notes sent after the call, which was cancelled due to a GTM glitch)
This will include single node and multi-node cases. The multinode case will use TOSCA relationships to establish inter-VM network connectivity. The examples will be very basic since they are intended to be a basis for further VNFD attribute tests.

The R3 planning is complete AFAIAC, but if anyone wants to get more specific re the JIRA items then feel free. Same for the Wiki - I think it's pretty good as a starting point, but contributions are welcome.

My minimum goal for R3 is as above, to establish a baseline for test development as part of the CI process, to augment the existing vIMS tests and further flesh out VNFD features as time permits. At least one VNFM (Cloudify) will be included, but support for others (JuJu, Rift.ware, ...) will be included given time and contributors. Since vIMS is (AFAICT) supported by all installers, a related goal will be that these tests will also be included in all CI runs. If that for some reason doesn't pan out (e.g. installer code freeze dates are too aggressive), the tests will be available at least as a post-install test suite.

2016-April-4 Models Project Meeting #4

Attendees:
  • Bryan Sullivan, AT&T
  •  
Agenda:
  • Status of project setup: Git, JIRA, Wiki
  • JIRA epics and story overview
  • Contributors to various epics/stories
  • R3 project plan and minimum goals
  • Events plan
  • Model analysis status
  • Tools analysis status
  • Model testing status

2016-March-21 Models Project Meeting #3b

Cancelled (no show)
See the Meeting #3a for minutes.

2016-March-21 Models Project Meeting #3a

Attendees:

  • Bryan Sullivan, AT&T
  • Serge Manning, Sprint
  • Lingli Deng, China Mobile

Agenda:

  • Status of project setup: Git, JIRA, Wiki
  • JIRA epics and story overview
  • Contributors to various epics/stories
  • R3 project plan and minimum goals
  • Events plan
  • Model analysis status
  • Tools analysis status
  • Model testing status

Meeting summary

  • The overall goal for the meeting was to share and get input on the project setup work so far and planning for R3
  • Most critical is the release plan: https://wiki.opnfv.org/models/releaseplan
    • MS0 needs to be completed by the end of March. The JIRA, Git, and wiki edits last week are intended to get us further toward that.
      • The resource estimate is intended to make it clear what time commitment is likely needed for the project to succeed, e.g. for Bryan a minimum 10% (4 hours/week, not counting meetings)
      • The JIRA epics/stories provide a starting point for specific items to make progress on
      • We can use the tasks/issues/etc associated with stories to plan detailed analysis steps and report progress
    • The minimum deliverables need to be agreed
      • docs: report (project goals and approach overview; architectural concept/backdrop for model-driven NFV analysis; SDO progress/convergence analysis; tools analysis; use case analysis; references)
      • use case tests
        • framework for model tests using various VNFM/NFVO open source software
        • specific use case tests to be implemented in R3
      • blueprints (as determined by analysis/tests)
  • the "Contributors" table at the bottom of the main wiki page is intended to help us focus on where we can add the most value; specific objectives
    • in general per area (SDOs, upstream projects, other mid-stream or OPNFV projects)
      • take lead on getting references to support/needs for info/data models and analysis related to them
      • publicizing the Models project and driving OPNFV engagement thru collaboration tools (eg. JIRA), e.g.
        • discussions on specific issues
        • ongoing alignment efforts (input/feedback, documenting consensus)
  • how to pick OPNFV and external projects to reach out to, from the list on the wiki
    • focus on overlap e.g. projects that need a medium for model-driven capabilities, or providing model capabilities
    • example is ODL; Bryan is looking into what YANG developers need and how YANG is used for service and device control
  • a model architecture would be good to help focus and explain our analysis of the role of models in VNF lifecycle mgmt
    • Bryan will draft a proposal
  • areas to focus on for 3GPP in the near term may include:
    • control/user plane separation for 4G (presumably different model components would be used for each)
    • also 5G with slicing disaggregation, etc
      • ON.Lab work on Mobile-CORD is relevant here
  • we'll continue on the list in the meantime and flesh out where we can contribute

Meeting ended at 15:59:29 UTC.

2016-March-7 Models Project Meeting #2b

Attendees:

  • Bryan Sullivan, AT&T
  • Ulas Kozat, Huawei
  • Prakash Ramachandran, Huawei
  • Uri Elzur, Intel

Open discussion of the project goals.

  • Prakash: suggest that the testing focus should be wider than CI/CD and Functest should come to own what we develop.
  • Bryan: that was the recommended plan
  • Ulas: intend to focus on SFC and continuous upgrades of VNFs and chain reconfig
  • Prakash: will also focus on scaling
  • Bryan: points to ONOS project proposal for Yang support: https://wiki.onosproject.org/display/ONOS/YANG+Models+in+ONOS
    • we can share the same model for testing across SDNCs to validate interop
  • Ulas: transportability of Tosca templates is a similar goal
  • Ulas: testing SFC in detail will require validation of each hop in the forwarding
  • Ulas: more complex test cases will require a formal model for the expectations so a test framework can validate it
  • Prakash: distribution of the templates is needed for policy that is applied across different domains/locations

We may take a 2nd look at the meeting time for the AM PST call; Ulas has a conflict at 8-9 AM and this is why he was able to attend this session.

2016-March-7 Models Project Meeting #2a

Attendees:

  • Bryan Sullivan, AT&T
  • Mark Flauw, HPE

We had issues getting the bridge to work, still need the logistics setup with the help of LF:

  • GTM bridge
  • Git repo
  • Jira project
  • IRC meetbot for #opnfv-models

The 2nd slot ('b") for the Models weekly meeting will be at 4PM PST Mondays. Bryan will find a GTM link we can borrow in the meantime until LF allocates one.

2016-Feb-29 Models Project Meeting #1

Attendees:

  • Bryan Sullivan, AT&T
  • Serge Manning, Sprint
  • Mark Flauw, HPE
  • Pierre Lynch, Ixia
  • Christopher Price, Ericsson
  • Pasi Vaananen, Stratus
  • Steve Furr, NXP

Agenda:

Project goals =

Open discussion and scoping (down) of the project goals for R3/C as needed.

Two main threads emerged in the call:

  • (1) SDO outreach and model convergence assessment/promotion;
  • (2) testing/demo of current model-driven design in current developer tools and platforms (including OPNFV Brahmaputra + extensions, and other adjunct tools e.g. orchestrators). Further clarification of these will take place thru Jira.

Project lead

In the meantime I can continue as project lead, but I would be fine with someone else stepping up to the role, since I'm already lead on one project (Copper).

Meeting format/cadence

Link to the doodle polls for the meeting times:

Bryan: Any day seems to work - we need to pick one, and to keep it simple I recommend Monday as well.

Tools: Jira, GIT, wiki, etc

  • a. Discuss how we will use these tools
    • Familiarity and regular use of the tools varies in the team. No specific suggestions so far on how to use the tools. We can just get engaged thru the tools and see who takes to what roles.
      Suggestions made by Bryan during the call include:
  • Jira for setting up the general structure of focus/tasks and SDO interaction, getting more specific as needed as we go.
  • Git/Gerrit for documents and other artifacts in development, pending determination of how familiar/comfortable the team members are with editing or submitting change requests thru "patching" documents etc. In the meantime, Bryan can help out as editor and incorporate suggestions thru calls/email/etherpad/etc into docs as needed.
  • Wiki and etherpad for general idea development and documentation. Rich media (e.g. Powerpoint, Gliffy, Word) can also be used and captured on the wiki prior to ingestion into docs as needed. Most important is collection of info/ideas with minimal impedance.
  • b. Assign tasks to set them up
    • As above, we will figure this out as we go.

Jira task setup

a. Types of docs we will develop
  • Specific suggestions are still welcome, beyond what the etherpad intro covered (doc structure TBD)
    • introductory docs with background clarifying the role of different types of models
    • inventory of references to standards-published models, and models published through other places
    • references to UML modeling from SDOs and related tools
    • normalizing model representations
    • derivation of models from open source implementations
    • references to open source implementations of network functions that could be used in PoCs/demos
    • use cases to be validated and test strategy for each
    • reports published for each release covering
      • Cross-project, SDO, and upstream collaboration
      • use cases, models, platform components, and standards validated
      • issues identified
    • requirements for upstream projects based upon prioritized gaps
  • Chris pointed out the importance of working with the Functest and other testing teams on the use case validation.
  • Bryan agreed clarifying the intent to develop a use case testing plan and strategy, and after some initial manual use case development, to work toward the goal of integrating the use case tests into the OPNFV functest library for ongoing automated testing.
b. Process for reaching establishing/maintaining SDO member participation/communication
  • Chris suggested to create a dialog with the SDOs on how we can consume their models and tools.
  • Serge mentioned that he is active in 3GPP SA2.
  • Marc mentioned that he is the rapporteur for ETSI NFV-IFA015 Report on Information Modeling, and active in 3GPP SA5. Further we could reach out for 3GPP SA5 help from Edwin Tse (Ericsson), Zhou Lan (Huawei), and Anatoly Andrianov (Nokia). Edwin had represented 3GPP at the NFV-ISG Information Modelling Workshop-Louisville: https://docbox.etsi.org/ISG/NFV/Open/Other/2016_01_13_NFV_Information_Modelling_Workshop-Louisville/
  • Marc mentioned that there is no expectation of a stage 3 doc from ETSI in this area;.
  • Bryan mentioned that Dan Druta will also be helping with ETSI NFV ISG interaction related to the IFA group, and that Serge and Marc in their SDO roles could help coordinate input/feedback to the SDOs.
c. Existing models and model-driven VNFdevelopment/management/orchestration open source projects
d. Use case development for model testing
e. Test environment for model testing
  • Bryan mentioned the goal to leverage community labs for the project, but in the meantime we need to find a lab environment for functional testing.
f. Expectations for gerrit reviews on issues of varioustypes (e.g. turnaround)
  • Bryan mentioned that unless we get into the mode of multiple direct contributors to docs etc thru git/gerrit (with gerrit-based reviews/comments/etc), he and the other committers can fast-track changes that are suggested and the group can review the results, suggest corrections etc.

Event planning

  • Discussion/hackfest/demo plans for ONS/OPNFV and other events e.g. OpenStack
    • OPNFV hackfest/meetup: https://etherpad.opnfv.org/p/Q1'2016_Hackfest
      • Monday 1:30 - 2:30: Test/demo of standardized data models for VNF management followed by Models project discussion
      • Hackfest/demos of tooling etc
  • ONS

 

(MS2) Detailed test case descriptions communicated to test project teams
  • No labels