Minutes IETF102: nfvrg

Meeting Minutes Network Function Virtualization (nfvrg) RG
Title Minutes IETF102: nfvrg
State Active
Other versions plain text
Last updated 2018-07-20

Meeting Minutes

NFVRG Meeting
Montreal, QC
--- Wednesday 18 July 2018, 1520-1650, Room Duluth ---
* Welcome and administrivia ~ 5 min
  + Scribe(s) and notetaker(s)
Note taker: Pedro Aranda
Jabber scribe: Evangelos Haleplidis

  + Agenda bashing
  + RG announcements
Diego: whether a particular aspect of network slicing or NS in general should
be part of NFV working items. Discussion on list Ramki: more traffic on the
list, please

* Update on current and proposed drafts ~ 10 min
  + draft-aranda-nfvrg-recursive-vnf (P. Aranda, IMDEA Networks)
Ramki (VMWare): Work ongoing on "network service mesh". You could take a look
Pedro: Abstractions are already there (LinkModels are abstract)
Ramki: So we could try to align them
>>>> RG adoption will go through the list <<<<

* Discovery Mechanisms in the MANO Stack (C. Bernardos, UC3M) ~ 20 min
Evangelos (Mojatatu Nets): Are you going to define a data model for this draft?
Security issues are relevant as well Carlos: Trying to reuse what is in
ETSI-NFV. If more is needed, we will develop  a new datamodel. Regarding
security, in multidomain we have already something, but in the context of fog,
something needs to be worked out. Laurent (Nokia): discovery considered, but
what about resource advertisement? Carlos: in the draft, all is based on
advertisements Laurent: do you consider a generic approach to advertisement,
other WGs are also doing this too Carlos: its good to look at other approaches
to come up with something more general Ramki: how do you manage separation of
concerns? who is going to manage the service? this is a relevant topic Carlos:
you definitely need to know who is in charge

* NFV research challenges related to network slicing ~ 55 min
  + NFV Slicing problems & impact analysis (A. Galis, University College
Reza (Nokia): NS is end-to-end and if you want to be pragmatic, not all
components will be VNFs. The real picture needs PNFs too. Alex: This picture
comes from ETSI. You are right... Reza: I can help Alex: there is too much
complexity Reza: multi-tenancy in 5G isn't what we have now, can take it
offline, but people talking about slicing should be taking into acclunt
hierarchical multi tenancy Alex: Agree, more in next presentation

  + NFV Slicing framework (L. Geng, ChinaMobile) - given by Alex,
Reza: another model missing: connectivity between NSO and Network Controller,
that is connectivity API Alex: agree, let's take it offlne Ramki: It is
important to take into account dymanic creation and lifecycle Alex: Yes. And
the question of the scale: we may well end considerong thousands of slices

  + Network Slicing Roles and Interfacing in NFV systems
    (P. Martinez-Julia, NICT)
Diego: NFVO is not right as "NFV Operator", be careful with overloading
well-known acronyms or definitions

  + NFV Gateway Function for Network Slicing (S. Homma, NTT)
Laurent: Tomorrow in NMRG slicing is also a topic... How to follow? Discuss in
NMRG, NFVRG, create WG in IETF? Shunsuke: I would like to discuss slicing in
IETF Alex (UCL): research questions for the next ten years... So that would be
part of IRTF. However, some questions can be discussed in a focused part in
IETF. We need to be pragmatic Reza: first thing is to define the API, which are
missing in picture. And, again, we need to consider mutitenancy as a basic goal
Alex: these are essential problems in slicing, yes Diego: remember: this is a
group around NFV, we should be focused on NFV matters