NFVRG IETF 94 Yokohama
Wednesday, 4 November 2015 Wednesday Morning Session (Room 502)
0900 Ð 1130
Chairs: Diego Lopez (diego.r.lopez@telefonica.com)
Ramki Krishnan (ramki_krishnan@dell.com)
Note takers: Pedro Aranda & Philip Eardley
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Welcome and administrative matters
Presenter: Ramki Krishnan
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-19.pdf
- 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/93/nfvrg.html
- Pedro Aranda and Philip Eardley for notes. Diego acted as jabber scribe
- List of NFVRG and NFV research-related events
- Interim meeting. Agenda still under development. It will soon be announced on the list
Structuring Live NFVRG documents
Presenter: Diego Lopez
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-14.pdf
- 19 live drafts, 3 adopted, more to come
- Organize them, because 19 to RFC clash with normal practices
- Shared in the list, mostly positive feedback
- Identify document classes and how to proceed with them
- One document per interest area and evolve them to RFC
- Discussion
o Carlos Bernardos: Agree with focusing but structure shouldnÕt restrict that much
¤ No problem with having more than 4 documents alive but moving beyond draft to RFC, example of gap analysis
o Qi Quwang: Visibility and orchestrationÉ visibility of what?
¤ Services and functions
o Dirk Kutscher: We should figure out what NFV has to do with evolution of internet technologies. E.g. service chaining, performance
¤ Could fit into second suggested new working direction
Update on Existing Drafts
Policy Architecture and Framework for NFV Infrastructures
draft-irtf-nfvrg-nfv-policy-arch
Presenter: Norival Figueira
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-1.pdf
- Discussion
o Diego: SUPA WG on policy management has just started. It needs stretching use cases. This could be ideal for them
Resource Management in Service Chaining
draft-irtf-nfvrg-resource-management-service-chain
Presenter: Seungik Lee
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-5.pdf
- Discussion
o Phillip Eardley: Timescales for optimization? (questions misunderstood)
¤ Two different strategies placement and ???É optimize the performance for resource usage in service chains
¤ Chair feedback: Timescale needs to be addressed in the draft
o Steven Wright: You conflated terms VM and VNFC
Verification of NFV Services: Problem Statement and Challenges
draft-irtf-nfvrg-service-verification
Presenter: Seungik Lee
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-2.pdf
- Discussion
o Ramki: Merger opportunities?
¤ No
Gap Analysis on Network Virtualization Activities
draft-bernardos-nfvrg-gaps-network-virtualization
Presenter: Carlos Bernardos
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-9.pdf
- On request for adoption as non-permanent document
- Discussion
o Diego: Non-permanent document but contributes to the core of the WG
Towards recursive virtualization and programming for network and cloud resources
draft-unify-nfvrg-recursive-programming
Presenter: Robert Szabo
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-17.pdf
- On request for adoption (area must be identified)
- Discussion
o Steven Wright: Recursion is important design goal and common API is a piece of that. Better to think of model-driven architecture than API.
¤ Yes, more generic view is possible
o Chairs: Discuss whether ÓrecursionÓ is the right term. Terminology on ÒjointÓ, ÒrecursiveÓ (should ÒuniformÓ or ÒhierarchicalÓ be used?)
o Diego
¤ Security issues through the Òjoint API"
¤ YANG everywhere?
¤ Suggest to bring it to ETSI NFV as an input for the IM/DM discussion
VNF Pool Orchestration for Automated Resiliency in Service Chains
draft-bernini-nfvrg-vnf-orchestration
Presenter: Pedro Aranda
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-4.pdf
- VNF pool as a key concept for orchestration and resource management in VNF chains
- Discussion
o Nicolas Bouthors: Do pools include physical resources
¤ Manage pools of VNFs in transparent way Ð nothing to do with controlling legacy. Not applicable to physical infrastructure
o Steven Wright: VNF pool is about robustness. How is this reflected in e2e service description?
¤ Included as part of description
DevOps for software-defined telecom Infrastructures
draft-unify-nfvrg-devops
Presenter: Catalin Meirosu
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-12.pdf
- On request for adoption (area must be identified)
o Very few hands shown when Steven Wright asked how many read it
- Suggest to bring ideas to possible DevOps application to I2NSF
- Discussion
o Diego: DevOps should be one of the hot topics affecting not the arch, but for operations
o Linda Dunbar: Read the draft, didnÕt understand how relates to NFV
¤ It covers how people use it in large data centers, we extend to smaller set-ups and identify topics for research.
o Steven Wright: Yes, itÕs about a problem statement. NFV is about chunks of software & continuous deployment. This is a different timescale to traditional software in telecoms network which is embedded in hardware. So testing and continuous integration
o Diego: Look at I2NFS, maybe DevOps principles could be applied there
Joint draft on Resource Orchestration
Presenter: Kostas Pentikousis
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-13.pdf
- Merge of unify-nfvrg-challenges and felix-nfvrg-recursive-orchestration
- Proposing a non-permanent document and interaction with other RGs/WGs
o To be at the datatracker during the coming 18-24 months
o Asking for additional contributions
- Use it as an example of what a non-permanent document should be
- Discussion
o Philip Eardley: Benefits across admin domains?
¤ Resources across admin domains can help. Use case in catastrophes: rolling blackouts when energy shortage. Coordinating this information with the orchestration to anticipate where to place VNFs
New Proposed Drafts
An Analysis of Container-based Platforms for NFV
draft-natarajan-nfvrg-containers-for-nfv
Presenter: Ramki Krishnan
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-11.pdf
- Discussion
o Dirk Kutscher: Doc is a nice start
o Nicolas Bouthors: Need to adapt network when container appears, people are playing with dynamically binding containers to ports (pre-configure) Does this bring issues?
¤ Security isolation is a key aspect needing research. Need to consider hybrid as VM gives isolation
Lightning talks introducing new work
draft-rorosz-nfvrg-vbaas (Robert Szabo)
draft-cai-nfvrg-recursive-monitor (Catalin Meirosu)
draft-bagnulo-nfvrg-topology (Marcelo Bagnulo)
draft-mlk-nfvrg-nfv-reliability-using-cots (Ramki Krishnan)
Slides:
https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-18.pdf
https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-7.pdf
https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-16.pdf
https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-3.pdf
- Discussion (all questions on MarceloÕs presentation on topologies)
o Charles Eckel: Do not understand why traffic will be different
¤ Current traffic patterns in DC is between servers (N-S << E-W traffics) In an NFV pop, the traffic that traverses the PoP is the one that makes the money. Thus O(N.S) == O(E.W)
o Steven Wright: Interesting topic.. two aspects. ETSI documents. placements in a core location make different traffic patterns that placing at an edge PoP. Related project in ONOS
o Diego: nice discussion for Heidelberg
o Kostas: Editorial. Not a problem statement yet, more considerations document. Design considerations. Retitle / collaboration: some can go to the challenges
¤ Here for input and possible ways forward
o David: Question is whether we want to reuse current PoPs or create a new arch for NFVI
Open-source and project updates
SFC architecture model using OpenFlow comparing using NSH
Presenter: Yuri Sekiya
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-15.pdf
- Discussion
o Uri Elzur: This discussion belongs in the SFC. NSH is data plane & SFC is control plane
o Dirk Kutscher: Not sure if the comparison is fair, but technical comparison
o Charles Eckel: This topic was part of the Hackathon. Look at github
The SUPERFLUIDITY project
Presenter: Pedro Aranda
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-8.pdf
- Discussion
o Nicolas Bouthors: What design pattern to simplify overall architecture and design?
¤ Aim is VNF components that are tiny and very flexible
A DevOps Toolkit for Networks
Presenter: Catalin Meirosu
Slides: https://www.ietf.org/proceedings/94/slides/slides-94-nfvrg-6.pdf
- Discussion
o Jamal Hadi: What would be transport underneath.
¤ 0mq
¤ What is this?
¤ Tcp, udp and inter-process comms.
¤ Scaling and robustness issues, are you fixing?
o Ramki: TPG would be a good one for service verification