1. ONAP Optimization Framework - HAS Shankar Narayanan - AT&T Labs Research snarayanan@research.att.com
08/15/2017

2. What is HAS ? (Internal name - Conductor)
A service that provides optimized Homing and resource allocation
Key features:
Supports policy driven homing
Policies are specified for each service as constraints for the homing problem
E.g., latency, availability of specific platform features, platform cost., etc
Allows deployment of services and VNFs on multi-site, multi-VIM infrastructure
Problem definition: Find optimal cloud/sites where the components of a service should be placed in, while respecting service constraints
Service Constraints
HAS
Optimal Homing solutions
Service components
Available Cloud/sites

3. Why HAS ?
Homing is a fundamental requirement for provisioning resources in ONAP
Plays a critical role in service creation workflows of SO:
Where to place VNFs for a customer request?
Which VNF instances to use for a customer?
How large of a VNF to allocate?
Assists service management workflows of VF-C/APP-C:
Where to increase capacity ?
Provide placement solutions for ONAP components
e.g., OOM invokes HAS to to get an optimal placement for DCAE micro-service to be in proximity to the VNF it is monitoring
Many of these problems require combining information from multiple systems, the search space is large, and constraints and objective functions of the problems are service/provider specific.

4. HAS in service instantiation workflows
Policy Framework
Service components
Fetch homing policies relevant to service
Constraints on VNFs (e.g., capacity )
Objective function (e.g., distance, cost)
INPUT:
VNFs to be homed
Order parameters (e.g. bandwidth)
Gateway
Request for homing
Get potential candidates
A&AI
* Get initial candidate set
Firewall
* Apply capacity constraints
Check capacity
SDNC
OUTPUT:
For each VNF either
a new location to spin up the VNF or
an existing service instance where the new order can be homed
* Apply policy constraints
* Optimize placement
DCAE
Service Orchestrator
HAS
Multi-VIM
Response call backs with Homing solutions
NOTE:
HAS is service-agnostic by design.
Service specific parameters/constraints are inputs/encoded in policy

5. Architectural Alignment of HAS
Homing requires data to be combined from multiple systems
Policy: Service/Provider specific constraints
AAI: Multi-site, multi-cloud inventory information
DCAE: Performance, utilization
SDNC: available capacity
Extensible set of optimizations
Multiple dimensions: platform may want to minimize cost, service may want to minimize distance
Objective functions can be service/provider specific
Extensible set of constraints
New services can compose more complex constraints using existing constraints
SDNC
A&AI
Policy
Extensible data request
Data Plugins A*
Greedy
Facility Location
Optimization Algorithm Plugins
Distance
Constraint Plugins
Diversity
Regulatory
H/W Features
Service Attributes

6. Discussion items Identify project dependencies and APIs
Northbound and Southbound SO
A&AI
Multi-VIM
SDNC
Identify use-cases
VoLTE, Residential Broadband, other (?)
Target Workflows – service deployment, other (?)
Identify and define policies for use-cases