USE CASE REALIZATION MEETING (Notes 2019-03-13)

BBS Use Case in R4.

1. A&AI SCHEMA CHANGE - (Resolved with BBS U/C, it will use an Edge to Logical Link instead of a A&AI Schema change)

2. PnP WIKI - Plug and Play Wiki needs to be updated. Update PnP-6230

3. REQUIREMENTS UPDATES - CP authenticated, triggers registration of PNF & tunnels. PNF sends confirmation notification/message. additional event from state change. Review VNFRQTS to see if there are updates. VNFRQTS - Vendor interfaces to Vendor products. 1. VNF 2. PNF. (none to 3rd party controllers). Focused on mainly how the VNFs should be orchestrated (HEAT templates, resources, #ports, IP, VM). From VNF RQTS intent what are the proper scripts to orchestrate and what are resources needed to assign to VNF. This is addressed by PNF registration. Identified gaps in PNF registration was implemented new requirements not specific to BBS.
4. 3RD PARTY CONTROLLERS - PNF requirements that are "must" requirements that don't make sense for a 3rd party controller. Section to specify the type of controller a PNF is using.
5. STANDARD FORMAT - BBS first U/C. "standard format of the flow", set of sequences & interaction to execute a U/C. Req insures ONAP can run on an ONAP structure. Domain specific controllers interacting w/ PNFs.

PNF Software Upgrade

Use Case (Scope Change)

Huawei is leading the new (re-scoped) PNF Software Upgrade with Ansible.

Created new Jira tickets in SDN-C. PTL Ajay has committed.

API defined, Req, Integration. Roll-back testing. Reuse of Casablanca Tests.

Platform Evolution for

Data Persistency

MariaDB and platform evolution for run-time persistent configuration information. Used in the OOF/PCI U/C. DCAE Post-gress DB. Primary inventory in A&AI. Base another database off of that. ConfigDB doesn't publish updates. PCI U/C if SDN-R knows of a config change it updates. SDN-R publishes changes on DMaaP. Network Maps & Topology & History. Timestamp. Controllers keeping state of network.

Focus is on making the FM Meta-Data GUI work first. PM dictionary "stretch goal".

Scope Change:

SDC-2095 - R4 5G U/C SDC: PM Dictionary GUI Display from PNF Onboarded Package Open

ETSI SOL001/SOL004 extension PNF

Modeling PNFD.

Mapping onboarding PNFD and ONAP "Platform" info/data model

Requirements work. Package requirements Section 7.2.

NFV meeting introduced not-backward compatible change in SOL004 2.6.1, adding delay to the project. Updates to the TOSCA Meta and the Manifest File, NFV CR NFVSOL(18)000746r3  CR review will be held Mar 14, 2019.

VNF-SDK Development Updates.

Developing a SO Service Instantiation Flow.

First flow was SDC distribution: ARCHCOM: InfoFlow - SDC Service Distribution

Discussion RAN Network Slicing Examples

Models of VNFs/PNFs. Simulators, descriptors for VNFs/PNFs and build a ONAP service.

DESCRIPTORS NEEDED - Using descriptors to onboard. Build a Service. Want PNFD/VNFD

modeling core, VNFs need to have descriptors for NSSF, NEF, NRF, PCF, UDM, AF, AUSF, AMF, SMF.

PNFDs for CU/DU. add descriptors in services that can wrap the descriptors. Add properties.

How slice impacts the resource. Need to characterize a service as a slice how do we when reserve the resources and to what granularity.

FILTERS - contribution from Borislav, ability to share services a network service want a nested service. Nested services can be dedicated or shared. Once composite service is instantiated the nested service is instantiated. Shared new instance isn't instantiated, instead orchestrator tries to find an existing instances. How to choose which existing instance (use criteria to pick). Filter on the composite service "I want to find a nested service with specific properties) e.g. latency < xyz.

NESTED SERVICE - Service A which inside/contained has Service B. Non-Recursive.  A "complex" vs a "Simple" service. Should network slices be "orchestrated"? Filter applied to SDC. Concept of Affinity. Coupling of resources vs sharing resources. CPU pinning, allocating all processing to the CPU. Capacity management. allow a higher level layers to do provisioning ONAP doesn't maintain the rules, the service itself maintains info. Policy Rules vs Rules associated with how the service is managed & scale & isolated. Slicing is a configuration that is a virtual overlay on existing components, maintaining that view of what the slice looks like & the associated resources.

A&AI and SLICING - A&AI would have bookkeeping of ANR different & non-overlapping. Slices where you cant share resources, what is "nailed up" and what is not. I can't touch these NF, but I can use those other NFs. A&AI Framework to track. Identify the available Resources. Modelling. A&AI, SDC, VIM, orchestration. Needs to know how to create slice-related services. Start with SDC. If creating a service, create all the associated resources (xNFs) allow OSS/BSS to take advantage of infrastructure. Don't maintain policies to track connections. To maintain slicing view this would impact components in ONAP. in SDC how do i create slices, resources assigned to slice and how to inventory a slice. E.g., EPC as a service, partitioning resources two or more EPCs. Non-overlapping this EPC#1 is only for IoT w/ this BW. EPC#2 for Video w/ High-BW. EPC#3 for Bronze subscribers for eMBB. Granularity of control down to the VIM level. What needs to be added to information model & descriptors. From ONAP perspective in A&AI 3 instances of vEPC service. Notion of a slice is at the OSS/BSS level: vEPC#1 mapps to IoT. ONAP orchestrates the virtual resources w/ the vEPC#1. A partition with a practical example.

M4 Milestone Requirements work for PNF PnP

1. REREGISTRATION STORY UPDATES - PnP Wiki updates at 5G - PNF Plug and Play#5G-PNFPlugandPlay-PNPREREGISTRATIONUSECASE

2. STEP 37a STATE CHANGE UPDATE -

State Change VES Event (type = State Change)

Old state & New State.

CPE Authentication Notification

BBS U/C is using this flow to "update" ONAP letting ONAP know that the PNF has been successfully been activated.

From VES Event Listener Document:

PIM Project - DMTF & Intel & E// - Physical Infrastructure Manager. Compose Physical resources on-demand as needed, or decompose & recompose as needed by NFVI flavors. VNF chain w/ K8S + OpenStack where VNFs are deployed wanted, instead of provisioning for peak, provision enough to start with.

Functions inside ONAP continue to enforce assured service levels for VNFs where network service determine that we're falling short, procure more resources.

An external application might collect capacity data to optimize what needs to be done based on the PIM layer + VIM. Build enough physical resources to do what's needed and virtualize them (in most cases) and go down to the VIM layer.

Physical containers on top of Bare Metal (where there is no VIM) just a different environment.

OpenStack/Blazer - would allocate more resources. K8S & OpenStack to enable RedFish APIs. Cross-NFVI invoked by PIM and once procured & registered resources wi/ given NFVI domain, OpenStack domain.

A&AI would then register these new resources. Ingest new infrastructure. A "Cloud Region" equiv to this smart-fluid infrastructure (can already model those).

Expected to be a R5 El Alto Use Case.

On-demand composition of PNFs. Break down PNF into parts. some PNFs are dynamically createable.

Describe what your physical resource looks like based on a TOSCA Template. Provision them as needed.

PIM - "pool" building (compute, memory) - Industry going to compose CPU, Memory + Disk; offerings from companies HP, E//, Intel et al; mostly ability t  compose higher order things. E.g. take compute resource + provision addition FPGA, memory, fabric.  "I want 3 cores, memory, disk".

PNF = pool of physical drawn; trays of resources, composed by Fabric for inter-connectivity. PNFaaS, PNFondemand.

Network Slicing, BBU pooling (VRAN2.0) Cloud RAN.

Capacity Optimization Module.

Delivery Model differences. google's delivery model = public cloud = VNFs; AWS (Amazon). Maintain pools of resources with direct control, disaggregated infrastructure.

USE CASE: The EDGE. Anywhere there is constraints in physical footprint; space. Capacity, this could manage and allot resources to different applications.

ONAP in Modeling. Model of infrastructure in DMTF, "Cloud"/containerized, physical infrastructure.

http://www.3gpp.org/DynaReport/38-series.htm

TMF models

Set of Modeling Layers. Bare metal at bottom of the stacks, PNF registered into Multi-Cloud/Multi-VIM.

VMware turns it into a VIM on-demand environment. Acceleration Management

ONAP Partner: RACK manageer (built into the hardware itself); RedFish: pt1: ability to replace APLI interfaces for H/W Control & Composition.