NFVRG IETF
97 Seoul
Tuesday 15
November - Grand Ballroom 1
15:50-18:20
Chaired by: Sarah Banks (sbanks@encrypted.net)
Dirk Kutscher (dku@dkutscher.net)
Note takers:
Dave Dolson, Alexandre Chataignon
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Welcome, administrative and general matters
Presenter: Sarah
Banks / Dirk Kutscher
Slides: https://www.ietf.org/proceedings/97/slides/slides-97-nfvrg-00-ipr-agenda-intro-05.pptx
-
Mailing
list: https://www.irtf.org/mailman/listinfo/nfvrg
-
Web site: http://trac.tools.ietf.org/group/irtf/trac/wiki/nfvrg
-
Proceedings: https://www.ietf.org/proceedings/97/nfvrg.html
-
Document
review request
-
OPNFV
presentation at the beginning of the meeting to adapt with HeatherŐs flight
schedule
-
Heads-up on
the ongoing discussions on network slicing at this IETF meeting
OPNFV Status Update
Presenter:
Heather Kirksey
Slides: https://www.ietf.org/proceedings/97/slides/slides-97-nfvrg-sessb-opnfv-update-00.pdf
Update on Adopted Drafts
draft-irtf-nfvrg-policy-based-resource-management
Presenter: I.
Jang
-
Discussion
o
Who read
this draft asks the chair.
¤ Not many. Please read. We need feedback.
draft-irtf-nfvrg-gaps-network-virtualization
Presenter: Carlos Bernardos
-
Discussion
o
Diego Lopez
(Jabber room): May be "network slicing" is a term associated with a
particular environment. I am thinking that a more general term regarding
resource or function sharing would be more adequate in the RG context
¤ Carlos: Network slicing may be overused term.
o
Chair: How
many folks have read the draft?
¤ Few
o
Carlos has
been quite happy with folks contributing text. Would probably ask for LC in Chicago
Update on Existing and Proposed Drafts
draft-bernardos-nfvrg-multidomain
Presenter: Carlos Bernardos
-
Discussion
o
Hannu Flinck (Nokia): Topology information exchanged over
which protocol?
¤ BGP LS being considered
o
Parviz Yegani (Huawei):
Is this intra- or inter-provider solution?
¤ Both
o
Parviz: Who owns the catalog?
¤ One catalog per provider
o
Parviz: Is this is a reference? We can get details?
Is this an architecture? Assurance
is to whom?
¤ To the customer, regardless of which SP they
are associated with
o
Parviz: You need to show top layer and other layers
¤ Slides show only a high-level view, there are
more details are in the draft. Please read and comment
o
Parviz: How does this differ from ETSI OSS
reference Os-Ma and IFA001?
¤ We are still looking at that
¤ Parviz: Customer has legacy OSS/bss
and new mano stack. Exchange between this an mano--ETSI is already working on this interface. Maybe you
can leverage work
¤ Sure, we want to leverage,
o
Diego Lopez
(Jabber room): I see this related to resource mgmt. Multi-domain scenario has
to do with resource sharing. Can this go to the earlier draft presented by
Jang? Could you work together?
¤ Agreed
draft-vazquez-nfvrg-netcod-function-virtualization
Presenter: Angeles Vazquez
-
Discussion
o
Chair: Who
read the draft?
¤ Very few
o
Chair: Do
you have any driving applications? What is the main idea? Improving
reliability? Higher level apps?
¤ It depends on which part of the network you
are, but in order to fix failures and reliability problems the usual practice
is over-provision, and this would be a solution for that
¤ Chair: it would be helpful to identify use cases.
o
Parviz Yegani (Huawei):
Clarification question: scope of work is in a VNF?
¤ It uses a VNF
¤ Parviz: You define new VNFs?
¤ Chair: We are overtime, this question to the
list please
o
Diego Lopez
(Jabber room): One of the ways people want to support service mobility and
scaling is by means of load balancers aware of instances of VNFs and
scheduling. Can we substitute need for load balancing by this architecture?
¤ That would be an advanced use of network
coding to code flows. Balance relative redundancy of flows.
o
Kyle Rose (Sandvine): What needs to change to put this in the
architecture? What is missing from infrastructure to allow this? Are you trying
to push standards?
¤ Some blocks of the framework do need to be
modified or extended.
Open Source Updates and Other Work Reports
Automated Resource Control in Virtualized
Network Environments
Presenter: P.
Martinez-Julia
-
Discussion
o
Chair: next
steps?
¤ Find a set of functions that a virtual
resource controller needs from lower layers, that are not currently
covered/provided by existing solution
o
Chair: are
you implementing this on some platform?
¤ We are early in the research phase, but we
plan to implement
VNF benchmarking
Presenter:
R. Rosa
Slides: https://www.ietf.org/proceedings/97/slides/slides-97-nfvrg-06-vnf-benchmarking-00.pdf
-
Discussion
o
Kohei Shiomoto (NTT); Is
this for in-service testing or for development state?
¤ Interesting question; do you consider the
service as a whole, or just the VNF?
o
?? (SKKU): Benchmarking
vs. perf. monitoring. Can it be used for performance monitoring?
¤ Yes. Benchmarking is one approach. Validation,
dimensioning could all co-exist, depending on the parameters you're playing
o
??: Are you
using your own MANO?
¤ Yes, from project UNIFY.
Optimal
Service Placement using Pseudo Service Chaining Mechanism
Presenter:
T. Na
-
Discussion
o
Andy Veitch
(Netcracker): under review by whom? Will it be
available in pre-release form?
¤ Maybe within a month.
o
Chair: Note
we shouldn't talk about RG documents as "IETF standards". You can say
specifications maybe
Introducing MEC
Presenter: Hannu Flinck
Slides: https://www.ietf.org/proceedings/97/slides/slides-97-nfvrg-08-introducing-mec-00.pdf
-
Disclaimer:
this is not an official ETSI presentation
-
Discussion
o
Dirk
Kutscher (Huawei): latency reduction is an active topic. What people have
noticed is that all apps benefit from low latency. So there are activities like
dcTcP, QUIC,.. Assuming
these get implemented, what use cases remain for MEC?
¤ Even using QUic, etc, do not resolve the issue of distance. MEC is keeping
critical apps hundreds of ms closer, leveraging
locality.
¤ Walter Haefner (Vodofone) LTE has ~20ms; in 5G we want ~1ms. distance is ~100km per ms.
o
Dirk
Kutscher: mobile throughput guidance. Do we have to make the server to be aware
of what happens in base station?
¤ To react to short-time-frame, TCP window
cannot see as much as the base station can see. Field trials show significant
improvement if the server can select window and application level coding
matching the radio conditions.
o
Dirk
Kutscher: next step: is this about MEC in non-3GPP networks?
¤ 5G are going to be multi-access. MEC gives
start in multi-access.