1. ONAP Optimization Framework (OOF) POC for Physical CellID (PCI) Optimization
August 21, 2018

2. ONAP Optimization Framework (OOF) and PCI
ONAP Optimization Framework (OOF) provides a policy-driven and model-driven framework for creating optimization applications for a broad range of use cases
For Casablanca, OOF enhancements include formulation and solving of optimization problems
We propose an OOF enhancement and Proof of Concept (POC) for Casablanca for Physical Cell ID (PCI) Optimization
PCI Optimization is a well-understood problem
The reading/writing of the Physical Cell ID (PCI) configuration parameter has been included in SDN-R POC
Objective of the POC is to demonstrate disaggregation of the PCI SON functionality, and the data flows needed to implement this in ONAP using OOF
We envision a new PCI-Handler Microservice to facilitate the data flow and trigger mechanism
The focus is on data flow and interfaces, and the role of OOF, PCI-Handler-MS, SDN-C Controller and Policy etc
Objective is to incorporate an algorithm in ONAP using OOF, not to develop an algorithm

3. ONAP Casablanca plan for OOF-PCI
Minimum viable product to make incremental progress in ONAP core modules
The POC will help to develop details of ONAP interfaces and flows needed:
(a) to get the required network data to OOF for optimization, 
(b) compute an optimization,
(c) apply policies, and
(c) execute it using the SDN-C Controller
Casablanca test cases involve:
New PCI-Handler-MS (on boarded on DCAE) – interfaces to OOF, Policy, DMaaP
OOF – interfaces to PCI-Handler-MS, Policy, SDN-C,
Policy – interface to PCI-Handler-MS, OOF, SDN-C
SDN-C – interface to RAN, OOF, Policy, DMaaP

4. ONAP Casablanca: PCI Optimization using OOF
REST API
DMaaP Message 7
(Netconf/Yang)
Config Value
Change Notifn 4b 4a 6
Simulated RAN
~2000 Cells
Nbr_list, PCI values
OOF
Policy
SDN-C 3 10 1d 1c 8
PCI Handler MS 9 11 1a
Config DB 1b 2 5
Config Value Change
(SDN-C work done in SDN-R team)
Step
Functionality
1a-1d
All modules loaded to support PCI 6
OOF gets PCI optimization policies from Policy 2
PCI-Handler MS fetches configuration policies from Policy 7
OOF queries SDN-C database to fetch data for cells in the region (REST API call) 3
Config change notification from RAN to SDN-C (e.g. Nbr list change) 8
OOF provides PCI Optimization result to PCI Handler MS (REST API call) 4a
SDN-C publishes config data change on DMaaP to PCI-Handler-MS. 9
PCI-Handler-MS provides PCI recommendation to Policy on DMaaP 4b
PCI-Handler MS obtains relevant info from SDN-C (REST API call) 10
Policy sends message to SDN-C with instruction for PCI configuration changes on DMaaP 5
PCI-Handler MS invokes OOF for pre-defined workflow for PCI Optimization (REST API call) 11
SDN-C applies config changes via Netconf

5. Flow / test case See latest flow and test case at:
ONAP OOF-based PCI wiki page
Intel Confidential

6. ONAP Optimization Framework – OOF (R3)
ONAP Optimization Framework (OOF) provides a declarative, policy-driven approach for creating and running a wide variety of optimization applications like Homing/Placement, PCI optimization, etc.
Policy Models
CMSO
HAS
PCI
SDNC
CM Portal simulator SO
PCI Handler MS
OOF
Route
CMSO
HAS
PCI*
API
OSDF
Policy
Actual policies
PCI Optimizer
CMSO Optimizer
Route Optimizer
Homing/Placement Optimizer (HAS)
Dependencies on other ONAP Projects
SDNC
A&AI
Multi Cloud
MUSIC
* See OOF-API spec at

7. PCI Handler MS architecture
Spring boot application
Core Logic
Upon notification from SDN-R, process it appropriately, including:
Waiting for more notifications
Buffering notifications if OOF has been triggered for same cluster
Checking if OOF has to be triggered, and triggering OOF …
Upon receiving response from OOF, trigger policy
Fetch and store policy updates
Core Logic
Database
REST Interface
DMaaP client
Other ONAP components
OOF
SDN-R
Policy
Note: This architecture depicts a standalone PCI micro-service. When this is moved to DCAE, the interfaces will be adapted appropriately.
Config information
Processed notifications (but not sent to OOF)
Buffered notifications
State information
RAN-Sim
Intel Confidential

8. SDN-R Config Database SDN-R has a RAN Config Database Attribute Format
MariaDB or MySQL
Query API exposed to other ONAP modules
Used by PCI-H-MS (step 4b) and OOF (step 7)
Put timestamp in API, and figure out history specs later
Attribute
Format
networkId
string
cellId
pciValue
integer
nbrList
list of cellId
lastModifiedTS
timestamp
API
Input
Output
Get cellList
networkId, ts
List of cellIds
Get PCI
cellId, ts
PCI Value
Get nbrList

9. Approach for RAN Simulator
Simulated Network
Yang model defined by SDN-R module will be used for simulation
ConfD* Basic Netconf Server will be used to simulate the RAN. C- API extension will be used to update the phy- cell-id values , to send ran-neighbor-list changes
Each netconf server will run as standalone process or Docker container
The netconf servers will be spawned by RANSim Controller based on the topology specified.
Python scripts and client will seed the initial configuration for each Node.
Each netconf server can send a mount request to SDN-R when they are spawned.
ran-neighbor-list update notification will be sent by netconf server to SDN-R based on the trigger from RANSim Controller
Will be able to accept phy-cell-id update trigger from SDN-R and forward same to RANSim Controller
ONAP
PCI MS
Policy
Web GUI
Robot Framework/
Python scripts
SDN-R
Simulated Network
Netconf server 1
ConfD
Extension
Yang Model
RANSim
Controller
Netconf server n
ConfD*
Extension
Yang Model
Config DB
Initial configuration
Python Script
*ConfD alternates are being evaluated

10. Approach for RAN Simulator
RANSim Controller
Spring Boot based micro service
RAN topology of ~2000 nodes to be simulated is defined into a Config DB
RAN Simulator spawns the Nodes based on this topology
Exposes following rest APIs to
Retrieve the current topology
Update phy-cell-id of a node to simulate collision or confusion
Drop a cell or create a new cell
Retrieve the new phy-cell-id set by SDN-R to a node
Start/Stop the network simulation
MariaDB will be used as the config database to store the topology
Robot scripts will drive the complete test sequences
Web GUI will show the current topology, phy- cell-id collision/confusions, ran-neighbor-list changes, updates from SDN-R, etc. Pictorial representation will be a stretch goal
ONAP
PCI MS
Policy
Web GUI
Robot Framework/
Python scripts
SDN-R
Simulated Network
Netconf server 1
ConfD
Extension
Yang Model
RANSim
Controller
Netconf server n
ConfD*
Extension
Yang Model
Config DB
Initial configuration
Python Script
*ConfD alternates are being evaluated

11. Modules Impacted New PCI-Handler-Microservice OOF
Develop microservice (onboard on to DCAE in next Dublin release)
Interfaces to OOF, Policy, DMaaP
Programmed to trigger PCI optimization request based on message from SDN-C (S-1b)
Fetches and applies configuration policies, which specify rules for processing input messages from SDN-C (S-2)
Gets config change message and additional info from SDN-C via DMaaP (S-4a,4b)
Requests PCI Optimization from OOF (S-5)
Receives PCI Optimization result, and sends recommendation to Policy (S-8,S-9)
OOF
Request/Response interface to provide PCI optimization results to PCI-Handler-MS (S-5, S-8)
Use new API in SDN-C to query database (S-7)
Modify existing interface to Policy to fetch PCI optimization policy (S-6)
Modify models to incorporate solver for optimization of PCI
Host solver algorithm (existing)

12. Modules Impacted Policy SDN-C (work done as part of SDN-R team)
Provide configuration policies to PCI-Handler-MS (S-2) – new interface
Modify existing interface to OOF to provide PCI optimization policies (S-6)
Send message to SDN-C with action request via DMaaP (S-10)
SDN-C (work done as part of SDN-R team)
Interface to RAN (simulated RAN for this project)
Modify RAN Information/Yang models to support PCI use case
Netconf interface to RAN: change PCI value (S-11), accept change notification (e.g. NbrList) (S-3)
Publish RAN config change on DMaaP (S-4a)
Database
Maintain database with RAN element / config data
Provide API for database query from other ONAP modules (S-4b,S-7) – could use DMaaP interface
Get and process action request message from Policy via DMaaP (S-10)
Track state of configuration before and after change applies (future support for Change Management)
Extend Directed Graphs to support PCI use case, as needed

13. Modules Impacted RAN Simulator Simulator program with about 2000 cells
Accurate representation of netconf interface to SDN-R
Include: Cellid, Nbr_List, PCI value, location (?)
Initial Nbr_list and PCI value computed offline and assigned
Ability to simulate change in Nbr_list to create potential for PCI conflict/confusion
Ability to send Nbr_list config change notification to SDN-C
Ability to accept PCI value config change from SDN-C
Test scripts to facilitate SDN-R / ONAP tests for use cases

14. Casablanca - Optimization Framework Enhancements
TOPIC
ICON
DESCRIPTION
Optimal placement of vNF
Placement of Mobility Virtual Network Elements (CUs) across the highly distributed edge clouds is a fundamental requirement. Service Providers must also optimize the performance of the 5G RAN in real-time.
Optimization problem formulation
Ability to model the problem as a constrained optimization problem, that is driven by policies – Potential use case examples: formulation of optimization problems at various levels: Customer (e.g. provisioning), Service (e.g. slice optimization ), Network (e.g. Routing, problems at the network planning level), Infrastructure (e.g. Placement) & Resource (e.g. License)
Optimization problem solving
Ability to use and deploy appropriate analytics, algorithms and solvers to solve the problem in acceptable time frames at various levels: Customer, Service (e.g. Slice Optimization Analytics), Network (e.g. SON network planning analytics), Infrastructure (e.g. Placement) & Resource
Impacted systems: SDC, SO, Policy, AAI, DCAE, SDN-R, MultiCloud (possibly)

15. Thank You!