Versions Compared

Old Version 1

changes.mady.by.user Jorge Hernandez

Saved on

compared with

New Version Current

changes.mady.by.user Sven van der Meer [VMware]

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

== Logistic

  • Date: 2017-07-12
  • Start: 1300 UTC+1
  • End: 1400 UTC+1

  • Lead: Jorge Fernandez
  • Minutes: Sven van der Meer

== Attendees

  • Ankit Patel
  • Avinash
  • Bobby Mander
  • Chenfei Gao
  • Guangrong
  • Huabing
  • John Keeney
  • Jorge Hernandez
  • LiZi
  • Liam Fallon
  • Lusheng Ji
  • MIchael Z
  • Mahesh Poojary
  • Manoj
  • Martial
  • Pasi Vaananen
  • Pramod Jamkhedka
  • Ruan
  • Ruijing
  • Sven van der Meer
  • Vijay Gopalakrishan
  • Yuan Liu
  • li n
  • lim
  • rgou
  • sarat
  • xinyuan

== Agenda

...

  • Contributor's time commitment to Policy Framework Project.
  • ONAP Optimization Framework Presentation by Ankit Patel.
Minutes:

...

  • The project is scheduled for the second release of ONAP.
  • Initially, configuration policies will be the only means of interaction with Policy to express constraints.
  • In the future, there may be usage of Drools for optimization processes.
  • The xcore modeling language is used to create models (JSON representation goes in the config policy).

...

  • TOP1: Contributor Time Commitment to Project
  • TOP2: SNIRO Presentation

== Discussion

=== TOP1

  • to all, please provide estimation of contribution to the project back to Pam
  • ONAP JIRA now has "Policy Board" with planning for Amsterdam release

=== TOP2


  • explains the flow (policy specification -> OF engine -> Graph algorithms/ILPs)
    • policies are read from the policy sub-system

  • Q:: is the policy provided by optimization or policy framework?
  • A: provided by policy framework but constraint by optimization framework

  • Q:: policy is currently for closed loop, how would that support an optimization policy?
  • A: policy framework provides several policy engines and languages (for domain), optimization policies are configuration policies and XACML is perfetly suitable for that
  • D: right now YAML is used a policy language (only operatinal policies at the moment), for optimization we need configuration which is different
  • D: configuration policies are freeform based on XACML with various payload, e.g. JSON
  • D: optimization might also introduce other solvers (FOSS) in its evolution, e.g. to use Drools as a language
  • D: Drools can also easily be translated into YAML


  • Q:: for release 2 (R2) only configuration? Drools later
  • A: yes (optimization is not part of R1)


  • Q:: The PDP is XACML?
  • A: XACML is used to store the policies, the optmization uses XACML to carry payload for optimization problems
  • D: intend is to convert (XACML) policy to an optimization problem definition


  • Q:: How does the optimization function work with policy?
  • Q: Which site is closer to the policy engine (framework) and which is closer to controllers
  • D: this project is not associated with policy framework, but requires it - but at the same time the optimization function can be used in a closed control loop as well
  • D: policy provides the input for what needs to be optimized
  • D:: would be good to provide some use cases to see how it works and flows
  • A: use cases
    • UC1: service instantiation (details discussed)
    • UC2: change management scheduling (details discussed)


  • Q:: what about the intended dynamicity and automation of the optimization
  • A: supports runtime optimization with online requests


  • Q: can this also be used for closed control loop use cases?
  • A: yes, e.g. optimize radio access network (maximize throughput using various control parameters)


  • Demo:: runtime and design time
    • showing Eclipse EMF with XCore
    • showing example policies in XCore and JSON
    • showing XMI export
    • showing policy editor with XMI import and policy creation (from AT&T E-Access portal)
    • showing deployment using PostMan


  • Q:: can you do the distance calculation (from demo) also in Drools
  • A: can be used OpenPlaner for optimization, it is using Drools, then we can integrate everything, so yes
  • A: but we haven't done that yet


  • Q:: the optimization is completely separated from the policy framework?
  • A : yes


  • Q:: can it be considered a contraint that is input to the optimization?
  • A: the contraints we have are modelled into policies, when policy is retrieved, the engine can translate that into contraints
  • A: by exposing contraint into policy 3rd party can define them, also everything is stored in policy repo and we can leverage other policy characteristics (e.g. prioritization, conflict)


  • Q:: how to plugin data?
  • A: policy requires data, so the framework provides the libraries to interface with data providers (runtime information)


  • Q:: can we put some sequence diagrams into the Wiki to understand the flow?
  • A: outside the scope of this presentation, but YES we should do that and provide more information on documentation
  • A: plan is to include more and more information, e.g. sequence diagrams

== Action Points

  • AP1:: (all): provide commitment to project to Pam and in confluence
  • AP2:: (Jorge, Pam): explain documentation plan and diagrams that are planned for Wiki

== Final Remarks

  • none