DRAFT PROPOSAL FOR COMMENTS

Overview

Project NameONAP Optimization Framework (OOF)
Target Release NameBeijing
Project Lifecycle StateIncubation
Participating Company AT&T, Intel, Netcracker, VMWare, ZTE (in lexical order)

Scope

What is this release trying to address?

The Beijing release is the first release of the OOF project and has the following key objectives in this release:

  1. Support ONAP use cases for Homing/Placement of VNFs and Services via the Homing and Allocation Service (HAS).
  2. Demonstrate the declarative, model-driven functionality of OSDF via representative applications aligned with ONAP use cases
  3. Provide a library of OSDF components/adapters for building new optimization applications

  4. Ensure compliance of OOF with platform maturity requirements. 

Additionally, the following stretch goals are also being pursued:

  1. Provide foundations of components required to support broader use cases that are targeted for R2+
  2. Exploratory efforts that are in the scope of near-term and longer-term OOF objectives

Detailed Scope Description 

ONAP Optimization Framework (OOF) provides the ability to create and run a policy-driven optimization application in a declarative, model-driven matter via the Optimization Service Design Framework (OSDF). The OSDF builds upon seed code from AT&T by bringing in new optimization technologies identified via collaborative efforts to provide a declarative, model-driven optimization framework. The OSDF also supports integration with custom coded optimization applications. This release includes the Homing and Allocation Service (HAS), which again builds upon seed code from AT&T, and addresses a core ONAP requirement, Homing/Placement of VNFs and Services (VNF Homing Capability).

Additional details for the initial objectives are as follows:

  1. Support minimum viable ONAP use cases for Homing/Placement of VNFs and Services via the Homing and Allocation Service
    Specific homing/placement use cases identified for this release are:
    1. Residential Broadband vCPE use case
    2. Hardware Platform Enablement (HPA) in ONAP
    3. VNF scale out use case
  2. Demonstrate the declarative, model-driven functionality of OSDF via representative applications aligned with ONAP use cases
    This will be done through a simple example mimicking Change Management Scheduling Optimization use case (this effort is aligned with the requirements of the broader CMSO use case). Additional examples are also being pursued, but are slated as stretch goals. The example will also demonstrate the ability to inject new types of constraints at run-time via policy (thus, in many cases, be able to inject advanced functionality into the optimization application without changing the model or configuration).
  3. Provide a library of OSDF components/adapters for building new optimization applications
    Beyond the adapters for Policy, AAI, MultiCloud, and SDC, additional focus will be on the incorporation of runtime metrics from MultiCloud for adaptive workload optimization
  4. Ensure compliance with platform maturity requirements
    This will be done by meeting or exceeding minimum requirements for projects in Beijing release.

The following items constitute “stretch goals” and “foundations towards real goals”

  1. Provide foundational components required to support broader use cases targeted for R2+
    The objective is build “foundations" of components that will be developed and be ready for integration with applications supporting broader use cases such as:
    1. VoLTE use case
    2. 5G family of use cases
    3. Broader CMSO use case
  2. Pursue exploratory efforts that are in the scope of near-term and longer-term OOF objectives
    We have identified some general features that we believe will be required in the near-term to longer term (timeframe or two or three releases in the future). However, the core OOF group will be focusing on the ONAP release deliverables and will not be able to spend substantial development effort on these items. Fortunately, these appear to align well with recent advances pursued by the broader optimization community. So, initially, we will pursue these via engaging the broader optimization community to contribute ideas and models. Specific features of interest currently are: 
    1. General algorithms for addressing uncertain data
    2. Robust scheduling under uncertainty
    3. Optimization of resource allocation when only approximate/averaged metrics are available

Use Cases

Minimum Viable Product

  • OOF-HAS - Homing Service (runtime) that can be provides optimized placement based on policy constraints, across multiple clouds and multiple sites
  • OOF-OSDF - Optimization Design framework (runtime) that supports HAS, the Homing Optimizer
  • Application demonstrating a declarative, model-driven, and policy-driven basic CMSO functionality (this effort is aligned with broader CMSO use case)

  • HPA functionality integration into OOF
  • Integration with Multi Cloud (incorporation of runtime metrics)

Functionalities

List the functionalities that this release is committing to deliver by providing a link to JIRA Epics and Stories. In the JIRA Priority field, specify the priority (either High, Medium, Low). The priority will be used in case de-scoping is required. Don't assign High priority to all functionalities.

Epics

Key Summary T Created Updated Due Assignee Reporter P Status Resolution
Loading...
Refresh



Stories

Key Summary T Created Updated Due Assignee Reporter P Status Resolution
Loading...
Refresh

Tasks and Sub-Tasks 

Key Summary T Created Updated Due Assignee Reporter P Status Resolution
Loading...
Refresh

Longer term roadmap


As the ONAP platform matures and new capabilities are introduced over time, the need for platform optimization services will grow along with it. The ONAP Optimization Framework is envisioned to handle this need as effectively as possible, by enabling creation of new optimization services with minimal or little new code development. The goal of OOF is to provide

(i) a growing set of core platform optimization services such as VNF placement and resource allocation (OOF-HAS), change management scheduling (OOF-CMSO), etc. 

(ii) a framework  (OOF-OSDF) that allow users to construct new optimization services and to extend/enhance existing platform optimization services.

Vision for OSDF:

OOF-OSDF is envisioned to be a collection of design time optimization libraries along with reusable runtime tools and microservices to facilitate and simplify the creation of new specific runtime optimization functionalities. The goal of OSDF is to avoid siloed optimization tools and associated duplicated efforts and overheads. For instance, the Homing Service, HAS (which is targeted for Beijing release) will not only contribute its reusable components to the framework, but also leverages the framework in its own feature development. Other potential optimization services that can be built using this framework include energy optimization in networks, optimal route selection for various network services, and radio access network (RAN) performance optimization. 

Vision for HAS:

OOF-HAS, or the Homing Service is a distributed resource broker that enables automated policy-driven optimized placement of services on a global heterogeneous platform using ONAP. HAS is architected as an extensible homing service that can accommodate a growing set of homing objectives, policy constraints, data sources and placement algorithms. It is service-agnostic by design and can easily onboard new services with minimal effort. HAS is designed to be used as a building block for both initial deployment, as well as runtime redeployment due to failures or runtime-capacity increase (scale-out). While the immediate deliverable of HAS in Beijing Release is to provide optimized homing/placement of services during the service instantiation workflows in ONAP,  HAS naturally extends to a general policy-driven optimizing placement platform for all platform placement functions, including placements of VMs, containers (e.g., for DCAE micro-services), ONAP Control Loops or VNF specific resources. HAS will also eventually allow placements of additional resource types such as licenses, VNF resources, etc.

Release Deliverables

Indicate the outcome (Executable, Source Code, Library, API description, Tool, Documentation, Release Note...) of this release.

Deliverable NameDeliverable Description
OOF-HAS

Executable, and source code for the Homing Service for the ONAP platform

OOF-OSDFExecutable and source code that provides optimization design framework support to optimizers like HAS; this will also include the simple CMSO application
OOF policy/data pluginsPlugins that help retrieve policies from the Policy framework (for homing, CMSO applications), and corresponding data required from AAI, MultiCloud, and SDC.
Service and API specification guides (homing and CMSO)Documenation for the following: (a) APIs exposed in R2 for different services, (b) how policies for different services can be created (e.g. OOF-HAS Homing Specification Guide, scheduling constraints, as well as general purpose declarative constraint specification).
TutorialsDocumentation on how a new type of policy- and model-driven application can be created by using declarative modeling, data templates, and minor configuration

Sub-Components

  • OOF-HAS - Homing Service that can be provides optimized placement based on policy constraints, across multiple clouds and multiple sites. 
  • OOF-OSDF - Optimization Design framework that helps create new types of policy- and model-driven applications. It also supports optimization modules that use custom code for solving the optimization problem. The OOF-OSDF also supports HAS, the Homing Optimizer

ONAP Dependencies

The following are the dependencies for the project based on the scope for the Beijing Release. The required dependencies have been identified based on the current homing requirements for the vCPE use case, and the potential dependencies are tentative dependencies that may exist based on how the information required for homing (e.g., Hardware Platform Enablement, VIM attributes) is available. 

oof-dependencies

Architecture

OOF Alignment with ONAP Target Architecture (Projects view):

Source: original image from 2018-01-09 version of the ONAP Beijing Architecture slides from the Architecture Tiger Team Contributions section in the Architecture Tiger Team wiki page. 


OOF-HAS High level architecture diagram:

Source: Slide from the presentation: FCAPS Telemetry - "Standardized Infrastructure Class statistics Model" (at the developer workshop: ONAP Project Developer Event: September 25 - 28, 2017, Saclay, France).

Sequence Diagram showing OOF interaction with other ONAP components during the Service Instantiation workflow of SO:

Platform Maturity

OOF is currently a non-MVP project, and is an incubation project in the Beijing release cycle (seed-code is upstreamed in this release). Therefore, some of the "Actual levels" for the platform maturity indicate a level 0, which is commensurate with the project's current incubation status. Our goal is to align with the Architecture committee's recommendation on using common tools/libraries for platform maturity, and therefore we will be dependent on the tools/libraries provided by these projects to achieve some of these targets. The Architecture Subcommittee's recommendations for the projects on achieving the platform maturity can be found here (ONAP Beijing Architecture Jan 9b.pptx). 

AreaActual LevelTargeted Level for current ReleaseHow, EvidencesComments
Performance0

1

OPTFRA-35 - Getting issue details... STATUS

  • Will need to work with the Benchmarking team to define baseline criteria for Optimization requests.
  • 0 -- none
  • 1 – baseline performance criteria identified and measured
  • 2 & 3 – performance improvement plans created & implemented
Stability01

OPTFRA-36 - Getting issue details... STATUS

  • 0 – none
  • 1 – 72 hours component level soak w/random transactions
  • 2 – 72 hours platform level soak w/random transactions
  • 3 – 6 months track record of reduced defect rate
Resiliency12

OPTFRA-37 - Getting issue details... STATUS

  • Seed-code natively supports manual failure and recovery. In R2, OOF components are stateless (or can rebuild cached state during runtime), and no rollbacks are required. This may change when reservations are supported in R3 and beyond.
  • Will need work with OOM/Music team to understand the tools/mechanisms they will be providing for Runtime projects for automated failure detection and recovery.
  • 0 – none
  • 1 – manual failure and recovery (< 30 minutes)
  • 2 – automated detection and recovery (single site)
  • 3 – automated detection and recovery (geo redundancy)
Security0

1

OPTFRA-38 - Getting issue details... STATUS

  • Will need help with CII badging.
  • 0 – none
  • 1 – CII Passing badge + 50% Test Coverage
  • 2 – CII Silver badge; internal communication encrypted; role-based access control and authorization for all calls
  • 3 – CII Gold
Scalability0

1


OPTFRA-39 - Getting issue details... STATUS

  • Need to work with OOM team to identify ways support for single site horizontal scaling.
  • 0 – no ability to scale
  • 1 – single site horizontal scaling
  • 2 – geographic scaling
  • 3 – scaling across multiple ONAP instances
Manageability0

1


OPTFRA-40 - Getting issue details... STATUS

  • Need to work with the Logging team for their recommendations. The two components of OOF (OSDF and HAS) needs to use the same logging system.
  • 1 – single logging system across components; instantiation in < 1 hour
  • 2 – ability to upgrade a single component; tracing across components; externalized configuration management
Usability11

OPTFRA-41 - Getting issue details... STATUS

  • 1 – user guide; deployment documentation; API documentation
  • 2 – UI consistency; usability testing; tutorial documentation

API Incoming Dependencies

List the API this project is expecting from other projects.
Prior to Release Planning review, Team Leads must agreed on the date by which the API will be fully defined. The API Delivery date must not be later than the release API Freeze date.

Prior to the delivery date, it is a good practice to organize an API review with the API consumers.

API NameAPI DescriptionAPI Definition DateAPI Delivery dateAPI Definition link (i.e.swagger)Status
PolicyPolicy Client API to create, update and retrieve homing policies.TBDTBD

work in progress, JIRA stories being created.

POLICY-56 - Getting issue details... STATUS

AAIREST Web Service provided by AAI, to query available cloud-regions, and existing service instances where a new order can be placed.Primarily dependent on the R1 REST API. Currently working with AAI on making HPA capabilities available.TBD

                      aai_rest_api_doc_v11.html


work in progress for HPA capabilities, JIRA stories being created
MultiCloudAPI to retrieve VIM capacities (infrastructure metrics model)TBDTBD
work in progress, JIRA stories being created
SDCAPI to retrieve license models and VNF requirements modelsTBDTBD
work in progress, JIRA stories being created

API Outgoing Dependencies

API this project is delivering to other projects.

API NameAPI DescriptionAPI Definition DateAPI Delivery dateAPI Definition link (i.e.swagger)Status
HAS APIAPI to submit homing requests, and retrieve homing solutions.TBDTBD
JIRA stories are being created in SO to replace R1 Homing stubs with the actual HAS API
Simple CMSO APIAPIs to submit change management scheduling optimization requests, and to retrieve solutions.TBDTBD
JIRA stories are being created for a mock CMSO calling agent, which will be deprecated in favor of the CMSO portal when integrating with the Change Management use case

Third Party Products Dependencies

Third Party Products mean products that are mandatory to provide services for your components. Development of new functionality in third party product may or not be expected.
The following link shows the dependencies and their license information: Project FOSS dependencies for OOF.

Testing and Integration Plans

The following testing will be ensured for OOF in this release:

  • Unit Tests and Code Coverage: We will use Python testing frameworks (combination of unittest, nose and coverage.py modules). We will work with our collaborators from MultiCloud who have used a similar ecosystem for their project.
  • Continuous System Integration Testing (CSIT): We will develop Robot Framework scripts for in Robot Framework.
  • Functional Test cases: OOF will be tested according to a small but representative test cases covering specific “concrete” optimization scenarios. These test cases will be documented in detail. 
  • End to End Test cases: We will adapt the Robot Framework scripts in consulation with the integration team.  

Gaps

This section is used to document a limitation on a functionality or platform support (i.e. limitations that we are currently aware, and required functionality will be delivered in a future Release). Since the project is in an incubation stage, we are focusing on bringing up the base functionality and then present information on limitations in a proper context. 

Gaps identifiedImpact
To fill outTo fill out

Known Defects and Issues

(Note: this is an incubation project with seed code being stabilized in ONAP; bug reports are on the way)

Key Summary T Created Updated Due Assignee Reporter P Status Resolution
Loading...
Refresh

Risks

List the risks identified for this release along with the plan to prevent the risk to occur (mitigation) and the plan of action in the case the risk would materialized (contingency).

Since the project is in an incubation stage, we are focusing on bringing up the base functionality and then present information on risks in a proper context.

Risk identifiedMitigation PlanContingency Plan
Availability of a global data store that will support resiliency and recovery (job queue).Use a cloud instance's central database (instead of a global database) to store information required for recovery. It is low risk because we anticipate Music or OOM's kubernetes etcd be available. We need strong consistency (and can tolerate some amount of lag)Since this will only affect optimization requests that are active/pending, we can ask the callers to resend the request if they receive unexpected error conditions
Availability of data to support use casesWork closely with the use case teams (e.g. vCPE, Scale out), data providers (e.g. A&AI, MultiCloud), and related projects (e.g. SO) to define the requirements more conservativelyDevelop the functionality but not activate the functionality of related constraints in the absence of data

Resources

Fill out the Resources Committed to the Release centralized page.

Release Milestone

The milestones are defined at the Release Level and all the supporting project agreed to comply with these dates.

Team Internal Milestone

This section is optional and may be used to document internal milestones within a project team or multiple project teams. For instance, in the case the team has made agreement with other team to deliver some artifacts on a certain date that are not in the release milestone, it is erecommended to provide these agreements and dates in this section.

It is not expected to have a detailed project plan.

DateProjectDeliverable
To fill outTo fill outTo fill out

Documentation, Training

  • Highlight the team contributions to the specific document related to he project (Config guide, installation guide...).
  • Highlight the team contributions to the overall Release Documentation and training asset
  • High level list of documentation, training and tutorials necessary to understand the release capabilities, configuration and operation.
  • Documentation includes items such as:
    • Installation instructions
    • Configuration instructions
    • Developer guide
    • End User guide
    • Admin guide
    • ...

Note

The Documentation project will provide the Documentation Tool Chain to edit, configure, store and publish all Documentation asset.


Other Information

Vendor Neutral

If this project is coming from an existing proprietary codebase, ensure that all proprietary trademarks, logos, product names, etc. have been removed. All ONAP deliverables must comply with this rule and be agnostic of any proprietary symbols.

Free and Open Source Software

FOSS activities are critical to the delivery of the whole ONAP initiative. The information may not be fully available at Release Planning, however to avoid late refactoring, it is critical to accomplish this task as early as possible.
List all third party Free and Open Source Software used within the release and provide License type (BSD, MIT, Apache, GNU GPL,... ).
In the case non Apache License are found inform immediately the TSC and the Release Manager and document your reasoning on why you believe we can use a non Apache version 2 license.

Each project must edit its project table available at Project FOSS.


Charter Compliance

The project team comply with the ONAP Charter.

3 Comments

  1. Shankaranarayanan Puzhavakath Narayanan ONAP must be platform-agnostic. For that reason, the  Usability area of the platform maturity matrix should include 'multi CPU architecture support.' That will guarantee that the services we produce under this project can be 'used' by ONAP instances running on platform-agnostic environments.

  2. Will the adapter for SDN-C included in the Beijing release? As mentioned in the page(Here), in the initial release, Data Adapter Library will include adapters to Policy, A&AI, SDN-C, Multi-Cloud, and SDC. However, it is not mentioned in the detailed description above.

    Thanks.

    1. Thanks Ruoyu. It is not part of a "core commitment" (minimal viable product commitment) for R2. So, I added it as a "stretch goal" for R2.