Hi,
Please find the draft versions of 1st and 2nd interim meeting minutes.
Couldn’t take detailed notes for the 1st meeting, as I dialed in while on the
road.
Kindly go through the notes and add/edit/change (with changes highlighted), so
that I could merge all the received changes.
Please do so by mid next week i.e. 14th Oct.
After that, I will upload to meeting proceedings.
Cheers
-sam
NVo3 Interim meeting Ð Oct 2nd 2014.
Attendees: Balaji, Bhumip, Dan R, David B, Don F, John S, Lothar R, Lucy Y,
Matthew B, Michael S, Osama, Raghu, Sebastein, Thomas N, Tom H, Zuquiang, Jesse
G, Paul Q, Ken G, Benson S, Sam A, Anoop G, Himanshu S, Alia A, KM, Larry K,
Nabil, Sharon, Linda D, Erik N,
Minutes:
Benson:
- All attendees will be muted to keep the noise to minimum. Unmute to ask a q.
- This is interim mtg, which is like any other meeting.
- Note well and all other rules apply.
- We have two drafts on agenda today.
Bhumip> Does IETF maintains the liaison to SDOÕs or the NVo3 WG?
Benson> IETF will maintain the liaison to other SDOÕs. That is what is meant in
the charter.
Lucy> TodayÕs agenda is not heavy. So, IÕd like to present, if Thomas N and
David B do not have a problem.
- IESG had a chance to look at the updated charter text.
Alia> IESG approved it today for internal review. Will take another 2 weeks for
external review. I have sent email to the mailing list to read the charter as
it is approved.
Alia> Do not expect much surprises from external review
Benson> We changed the text around the control plane. Changed to one protocol.
Second one is adding LISP at the end along with BGP. i.e. work of these will be
taken in their respective WG.
Alia> NVo3 to focus on orchestration and not be like routing protocol like.
This is to make quick progress.
Benson> If I understand, one protocol per each aspect and not have just one
protocol for everything.
Fabio> What is the way the three WGÕs could talk?
Benson> Participants are overlapping in each of the WG. As you said, Individual
contributions will help co-ordinate.
GUE draft: (Tom H)
- This is the protocol we actually deployed. Lucy provided contributions to the
text
- There are several requirements related to control, security and performance
Bhumip> Is it for DC only or generic
Tom> It could used for other cases but have to consider somethings if want to
use it for other cases
- Need to integrate with switches, middle boxes.
- It should integrate with security protocols
- In use cases, encapsulation should support IPv4 and v6
- In virtualization use case, jobs could be migratable. Performance is critical.
- Third party VM;s could be integrated. Controling the packets is vital.
Bhumip> IS this similar to SDDC?
Tom H> This is lower layer primarily dealing with scheduling. For ex: prempting
higher priority jobs for lower priority, which could be similar to SDDC
- This encap will be done in IP/UDP
- GUE metaheader is the preamble to all the udp encaps
- Provided details about various fields in the header.
Bhumip> Header virtual or not? Can you have multiple instances of header?
Tom> You can have multiple layers.
David Black> IT is not sufficient to say Ômay set checksum to 0Õ
Tom H> You do realize that VXLAN differs from that.
David B> VXLAN is not ietf protocol or standard.
Tom H> What do you suggest
David B> We will have some text for MPLSoUDP draft. Hope that will help.
Alia> By the time this draft progresses, we should have some specific related
to mplsoudp draft right?
David B> By Honolulu, we should have text for MPLSoUDP. We could take it from
there.
Erik> Can you do checksum offload?
TomH> Little different from tx and rx
- Packet originated on VN should stay on VN. Security is important for this.
- Congestion control is needed. Lighter weight DCCP for CC.
David B> There is circuit breaker draft. Please look at it to use the right
terminology.
Tom H> I looked at it and in this context, I did mean CC.
Sharon> We have lot of experience using DCCP. We can take offline and share the
experiences.
Anoop> Is the CC operating on outer header or individual flows?
Tom H> CC of the tunnel
- foo o UDP in upstream linux
- Remote checksum offload is in development
Erik> You want this to be implementable?
Tom H> We cannot wait for h/w to be supporting this. But have this protocol
implemented.
Erik> Have you looked at supported by existing silicon?
TomH> That is interesting. One vendor said, yes.
Hpvr2NVE Ð (Lucy) :
- No presentation at this interim
Open Discussion:
Benson> Take a look at this BIER BOF and its scope. Will send out the like
http://tools.ietf.org/bof/trac/%23Routing
Alia> It is imp for WG to take a look at this BOF and its impact/influence on
NVo3.
Fabio> NVE to NVA is part of the scope. Reading from the draft, not sure NVo3
is focusing on.
Benson> We need to clarify what it means by orchestration.
Alia> We know this could be solved in many ways. Recharter will ack multiple
mechanisms to exist.
Fabio> If we leave multiple solutions, not sure if that is a progress.
Benson> NVo3 does need to solve and need to learn from other WGs
Fabio> Agree with Benson. For ex: How do you push the mapping from NVE to NVA
Bhumip> May be last paragraph need to be removed as it is causing the problem
Alia> It is needed. We had more than 2 years to discuss about this.
Osama> Can we use BGP between NVE and NVA? Is that in scope or out of scope?
Benson> Out of scope of NVo3.
Lucy> Why we use MAY for requirements and Use case?
Alia> Want to do parallel and not waterfall model
Lucy> I agree but we already have the docs done, hence the question
Erik> Happy that we focused and chartered to get work done.
Benson> there are many CP exist. Learn from them and start using them. We
should start discussing them on the mailing list.
Benson> See you all in the next interim.
NVo3 Interim meeting: 09/11/14
Attendees: Benson S, Dan R, Don F, Eric N, Jesse G, Lizhong J, Lucy Y, Marc L,
Alia A, Katherine Z, Lawrence K, Matthew B, Sharon, Thomas N, Tim H, Qiang Z,
Ken G, Tom H, Michael S, Fabio M, Jon H, Sridhar T,
Minutes:
General Discussion:
- Everyone should be on mute when they join.
- Welcome to first of three interim meeting
- The meeting is recorded and will be shared with the group
- Various logistics were discussed to better handle the virtual meeting.
- Linda has a scheduling conflict, so, may not have her presentation in todays
schedule.
#1 Charter Discussion:
- WGLC was made for the updated charter.
- Received various comments. Few of important comments will be addressed
Comment 1:
- Received comments from Thomas to remove certain terms like optimization,
management etc from para #1.
- Proposed resolution is to remove ÔoptimizationÕ from the charter.
- Lucy gave her ack for the change.
Comment 2:
- Clarify Logically centralized
- Benson proposed that we should rely on PS, FW, etc. for clarification, but
leave the text as it is.
Comment 3:
- Clarification on use of end device and NVE
- Proposed resolution is to leave as it is. There could be some ambiguity but
it is better to leave like that.
Thomas> For the point #1, NVE has to do management and security but we are not
doing. As it creates unwanted attention, want it to be removed. But better to
say, it will address these topics.
Benson> Will talk it through with Alia.
Alia> Makes sense to me to.
Thomas> Why are we using end device instead of host
Benson> End device refers to physical device.
Lucy> We could clarify in the arch doc.
- There was lot of discussion of end-device NVE control plane protocol is for
split NVE protocol.
- Suggested that, at this stage not to narrow down the choices but keep the
broader sense for scope of the work, which could be done.
Thomas> concern is about the arch, which will have impact on the chartered
items.
Thomas> End device NVE protocol should be option as it only applies in split
NVE case.
- The discussion is about to work on protocol or describe mechanism.
- Benson proposed that Ôprotocols should fulfill the mechanism for end device
NVEÕ.
- Benson to post the charter with updates and Alia could take it up with IESG.
- Benson and Alia have advised WG participate and get the work done and use the
interim meetings to speed up the work.
- Chartered items with milestones do not correspond to a draft.
Lucy> Is OAM and management the same or will have separate milestone?
Benson> Will take a look at it.
#2 Use Case Ð Lucy:
There are three use cases described in this doc. Provided details about each
use case.
Want to make WG LC soon.
Tom H> Where does the use cases for single tenant but multiple virtual networks
fall?
Lucy> VN within same tenant could talk, but different tenants VN doesnÕt.
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3
