Modeling (5/11/17)

I am wondering how modeling project will model the NFVI resources, while I found only "Resource models (that contain VF and VFC)".

My concern is: Will modeling project  cover the information model of NFVI resources? This is very critical to enable VF/VFC verndors and service providers to leverage the NFVI resources like EPA features in an automation approach.

One possible scenario is that when the VF/VFC is implemented assuming there are some type EPA features exist, Vendor should make it  specified in some specification by VNF-SDK,  and Multi-VIM will report those EPA resources to Catalog and AAI, so the SDC user could design the service based on these VF/VFC, operator could deploy such service later.

Without resource model to describe the catalog and quantity of EPA capability/resources, Both VF/VFC vendors and Service Providers will have to align to each other by offline approach, which is a deviation of automation.

I don't think Modeling could dictating other code projects, code we have is only about parsers and translators, for vnf/ns spec, modeling project need to work together with VNF SDK/SO/VF-C et al., for information model, we need to work together with A&AI project as well. Please help to join our contributor list if you would. thanks

Chat from 16-MAY-2017 call:

From Bryan Sullivan to Everyone: 06:38 AM
I disagree in principle that declarative methods can'rt handle complex use cases.
From DENG Hui to Everyone: 06:39 AM
Bryan, do you have example, how vendor's volte could be deployed based on declarative methods?
From Bryan Sullivan to Everyone: 06:39 AM
Dependencies as an example is clearly supported by a declarative model, and can be used to derive VNFC sequencing.
From Alex Vul to Everyone: 06:40 AM
I agree with Bryan...
From Bryan Sullivan to Everyone: 06:40 AM
You need to list the particular aspects of the model (actions in a deployment process for example) that need to be assessed per support in declarative means. Do we have an IRC channel for the meeting? Will be more effective than this chat tool.
From Alex Vul to Everyone: 06:42 AM
just look at the VMware proprietary implementaton or the BMC one, or examples of complex SAP landscapes that have been modelled in TOSCA the complexity of SAP landscapes rivals what we need to do in terms of network services, VNFs and SFCs
From Bryan Sullivan to Everyone: 06:42 AM
Also, the Cloudify DSL has a lot of complex features that they have extended TOSCA with. These have not worked into the standard but are a clear example of how to declaratively model complex features.
From Alex Vul to Everyone: 06:43 AM
I have been working with declarative orchestrators for more than 10 years - they can do the job...
From maopeng to Everyone: 06:44 AM
could the Tosca deal with both deploy and upgrade?
From Bryan Sullivan to Everyone: 06:44 AM
Same for JuJu, which uses a largely declarative approach, supplemented by scripts where needed for any detailed actions not yet supported by the model DSL. This is a common technique, and is a design decision in the end (how much to script based upon lifecycle hook calls).
From DENG Hui to Everyone: 06:45 AM
one more clarification again: is there any vendor's VOLTE could be deployed based on declarative model? please show the example,
From Bryan Sullivan to Everyone: 06:45 AM
I see no reason that upgrade cannot be represented by a reference to a dependent artifact that has changed version (upgrade or downgrade), based upon the current running state of a VNF.
From DENG Hui to Everyone: 06:45 AM
today's svnfm is under JUJU,
From Alex Vul to Everyone: 06:45 AM
TOSCA *is* for deployment
From maopeng to Everyone: 06:46 AM
@Alex?just for deployment?
From Bryan Sullivan to Everyone: 06:46 AM
Deng, the "vendor's VOLTE" example needs to be defined re requirements by the vendor. We can't in a generic sense respond to a particular example withouth understanding the assumptions/requirements of that example.
From Alex Vul to Everyone: 06:46 AM
you can also orchestrate upgrades/updates
From Amir Levy to Everyone: 06:47 AM
@DENG - we have ran VoLTE over declarative model using Cloudify
From Alex Vul to Everyone: 06:47 AM
the containment graph is helpful for in-service upgrades
From Amir Levy to Everyone: 06:47 AM
https://www.youtube.com/watch?v=NfkCkj3Hd8U
From DENG Hui to Everyone: 06:47 AM
@amir, which operator , and which telecom vendor, don't tell me that is open source
From maopeng to Everyone: 06:48 AM
one Tosca file or two files do deploy and update?
From Amir Levy to Everyone: 06:48 AM
@DENG check out - OPNFV / ClearWater
From Bryan Sullivan to Everyone: 06:48 AM
Deng, where possible we should always use open source examples to develop ONAP functionality. (if I understood your point/question)
From DENG Hui to Everyone: 06:49 AM
Bryan, we are talking about commerical value and commerical deployment, not just for academy based on open source vnf
From Bryan Sullivan to Everyone: 06:51 AM
If we can't derive clear deployable platform value from developing a platform using open source reference VNFs, then we might as well go home now. That does not impact your ability to deploy proprietary "real" VNFs, or diminish the potential quality of the platform to support them. We just have to do a thorough job using open source tools including VNFs.
From maopeng to Everyone: 06:51 AM
if the deploy and the update have the different relations or other depandencies running state, how to do that?
From Alex Vul to Everyone: 06:52 AM
can you elaborate maopeng...
From Bryan Sullivan to Everyone: 06:52 AM
We are not in a science experiment here; we are working in the open on a real deployable platform. We just need to keep all the components of the project discussion open and free.
@alex, we can discuss offline
From Me to Everyone: 06:54 AM
Nokia successfully deployed a commercial IMS using a TOSCA driven templates decribed here:https://tools.ext.nokia.com/asset/200827
From Alex Vul to Everyone: 06:55 AM
sure... i spent 15 years working on app modeling... what Huabing is proposing is actually one implementation that I have done...
From Bryan Sullivan to Everyone: 06:55 AM
Do these chat notes get published in the minutes? If not we need to start using IRC so the notes are minuted.

Some review comments on the project proposal:

1. For the Architectural Alignment, we should consider embedding an image of the ONAP Architecture diagram with the Modeling highlighted, per New Project Proposals (bottom of the page).
2. We should create a Committer list of resources and a Contributor list of resources.
3. The Primary Contact and Project Lead should be identified.
4. The Lifecycle State should be Incubation
5. There is currently one Repo name identified, should we consider multiple Repos, e.g., Information Models (with dimensions of Resource, Service, Workflow, etc.), Tooling, Processes/Guidelines, Data Models, Workflows
6. We might want to consider adding a What Problem is this project trying to Solve
7. Regarding these statements:

Code - parsers (unified API interfaces) and tools

What does this mean?  Wouldn't the individual ONAP projects be responsible for contributing their project specific Information Models, Data Models, Process Workflows and API definitions per the Modeling framework, Modeling guidelines and Common Information Model?

• Support Use cases for R1: vCPE, vVoLTE, vFW/vDNS, example DMs for this use cases

Should we consider creating not only example Data Models but also example Common Information Model and Process Workflows for the Release 1 Use Case(s)?  This will provide an example for all the other projects to understand what Modeling is all about and why the project exists.

link to IM and DM from OPENECOMP and OPENO

The links provided are to YAML data models, there are no Information Models at those links.

One review comment on the modeling project proposal description.  There is published document from ETSI NFV which describes the touchpoints/relations between the NFV Information Model and information models from other organisations, such as ONF, 3GPP, TMF. So, I am proposing adding new bullet point.

• How does this align with external standards/specifications?
• Oasis Tosca NFV profile
• ETSI NFV SOL spec
• ETSI IFA024: Report on External Touchpoints related to NFV Information Model

Scope of this project is overlapping with all other ONAP projects as individual projects, based on their functional requirements shall be defining their data model and information model. However since the Modeling project intends to unify all the data and information models across ONAP projects to harmonize all these models and create "Common Data Model" and consistent Information Model, handshake will be required with all projects and for that we need to devise a process (Process document) capturing,

• Acceptable formats for publishing data models by individual projects. It should include individual data items, relationships, cardinality, usage, description, Reference to ETSI NFV standard if any etc.
• Modeling team will collect, comprehend, collaborate and consolidate the models received from individual projects and propose updated data model back to individual projects with justification for each change. While making the changes Modeling team will ensure that changes in the proposed model does not break any of the intended function of the project. 
• Format for the details of public Interfaces/APIs exposed towards each ONAP module.
• Format for the details of expected/required Interfaces/APIs from each ONAP module.

The proposed process document shall help in

• Providing clarity and visibility to individual project teams on modeling items.
• Faster collaboration to arrive at common unified model.
• Creation of detailed workflows
• Integration of all ONAP components

NOTE: I am NOT trying to say this project won't work. I AM trying to say I have timeline concerns.

"Time plan:

Info Model specification finalize at 1.5 month before functional freeze(for R1, 20th June, 2017)
Data Model specification finalize at 1 month before functional freeze (for R1, 3rd, July, 2017)"

I'm hoping that the above are typos, or that they mean yes, here are the **requirements** to be finalized for our models. To me, "model specification" means that the technical specification for the model is done.

I do not see any info model right now (please correct me if there is one), and do not think that it is possible to build one in < 3 weeks. (I've been building info and data models for well over 30 years).

If there IS an info model, then I would like to see it. I am the Orchestration Area Director for the MEF and a TMF Distinguished Fellow. I would like to see what is being reused vs. modified vs. created, for these and other critical modeling sources (e.g., ONF, NFV).

If there IS an info model, then I do not think that you could generate a generic set of reusable and extensible data models in 2 weeks. I am sure that one or more data models currently exist, but they likely exist as bottom-up implementations that may or may not be related to each other. Specifically, I doubt that two different data models were developed from a common information model, because if they were, this is a huge selling point and should be able to be seen somewhere! If I am correct, then you would have to "backwards-engineer" the data models to come up with an info model, and that is exactly the wrong way to build an info model! For example, suppose I had an RDBMS, and implemented an object using tables related to each other by primary and foreign keys. This would prompt me to use FKs/PKs as object identifiers. Now suppose I want a directory data model. It doesn't know what a PK or FK is. So I now have to translate relational calculus concepts to constructs appropriate for the new data model. Furthermore, I likely have to rework my queries, since LDAP (or X.500) is optimized for text, while SQL is much more expressive and powerful. This becomes increasingly difficult from a bottom-up approach because each data model relies on making decisions for a particular platform, language, and protocol; info models should be independent of all of these.

Looking at the MEF wiki there seems to be some good pages that explain the terms: Information Model and Data Model

https://wiki.mef.net/display/CESG/MEF+Common+Information+Model

https://wiki.mef.net/display/CESG/Data+Model

Do this team agree that these pages are relevant for how ONAP understand the terms?

Where can I find the ZOOM recordings of the Modeling meetings?