Somesh, 

You are correct that this description does imply that TOR has physical links to 
the server which host the VMs. 

We will find out a way to change the wording to include the Virtual links case. 

Thanks, Linda

> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of
> Somesh Gupta
> Sent: Friday, August 24, 2012 4:28 PM
> To: [email protected]
> Subject: Re: [nvo3] Comments on Live Migration and VLAN-IDs
> 
> I see the following in the draft
> 
> "We say that an L2 site contains a given VM (or that a given VM is in
> a given L2 site), if the server presently hosting this VM is
> connected to a ToR that is part of that site."
> 
> and it seems to be at odds with "L2 site" being a virtual link
> in the hypervisor. Or maybe it is just my ability to understand :-)
> 
> Somesh
> 
> > -----Original Message-----
> > From: [email protected] [mailto:[email protected]] On Behalf
> Of
> > Linda Dunbar
> > Sent: Friday, August 24, 2012 2:11 PM
> > To: Black, David; [email protected]
> > Subject: Re: [nvo3] Comments on Live Migration and VLAN-IDs
> >
> > Agree with David on the meaning of L2 site. If you read carefully the
> > definition of "L2 site" in the draft (instead of various connotations
> > of L2 site under other context), the definition is very precise and
> > correct for the usage of this document.
> >
> >
> > Agree with David that "If every "L2 site" is a single virtual link
> > within a hypervisor, that requirement is rather easy to meet,
> > especially for virtual links that aren't VLAN-tagged ;-)."
> >
> > But most, if not all, data centers today don't have the Hypervisors
> > which can encapsulate the NVo3 defined header. The deployment to all
> > 100% NVo3 header based servers won't happen overnight. One thing for
> > sure that you will see data centers with mixed types of servers for
> > very long time.
> >
> >
> > If NVEs are in the ToR, you will see mixed scenario of blade servers,
> > servers with simple virtual switches, or even IEEE802.1Qbg's VEPA. So
> > it is necessary for NVo3 to deal with the "L2 Site" defined in this
> > draft.
> >
> > Linda Dunbar
> >
> > > -----Original Message-----
> > > From: [email protected] [mailto:[email protected]] On
> Behalf
> > Of
> > > Black, David
> > > Sent: Wednesday, August 22, 2012 2:20 PM
> > > To: [email protected]
> > > Subject: Re: [nvo3] Comments on Live Migration and VLAN-IDs
> > >
> > > > It is not clear to me why an L2-CUG would be tied to a VLAN-ID
> > > > in an nvo3 network?
> > >
> > > Uhm, it's not, but the draft could be much clearer on this topic.
> > >
> > > IMHO, the draft uses the term "L2 site" in a rather non-intuitive
> > > fashion for nvo3.  Here's the definition from Section 2 of the
> draft:
> > >
> > >    In this document the term "L2 site" refers to a collection of
> > >    interconnected devices that perform forwarding based on the
> > >    information carried in the Ethernet header. In a non-trivial L2
> > site
> > >    (site that contains multiple forwarding entities) forwarding
> could
> > > be
> > >    provided by such layer 2 technologies as Spanning Tree Protocol
> > >    (STP), etc...  Note that any multi-chassis LAG is treated as a
> > > single
> > >    L2 site - a given multi-chassis LAG has to be contained within a
> > >    single L2 site.  This document assumes that a layer 2 access
> > domain
> > >    is an L2 site.
> > >
> > > It's easy to mis-read that definition and infer than an "L2 site"
> is
> > a
> > > data center, but that's not what's going on.  For example, consider
> > the
> > > scenario in which the NVEs are located in the hypervisors (and
> > possibly
> > > also the DCBRs) - the "L2 site" adjacent to a VM then consists of
> one
> > > or
> > > more virtual links in the hypervisor, and none of the physical
> > network
> > > components in the data center are part of any "L2 site" (as that
> term
> > > is defined in the text quoted above).  Among the interesting
> > > consequences
> > > for this scenario is that the following requirement (from Section
> 3.1
> > > of
> > > the draft) is far less restrictive than it may initially appear:
> > >
> > >    This document assumes that VMs that belong to the same L2-based
> > CUG,
> > >    and are in the same L2 site MUST use the same VLAN-ID.
> > >
> > > If every "L2 site" is a single virtual link within a hypervisor,
> that
> > > requirement is rather easy to meet, especially for virtual links
> that
> > > aren't VLAN-tagged ;-).
> > >
> > > When the NVEs are in the ToR switches, the "L2 site" may be as
> small
> > as
> > > a single Ethernet link between the physical server and the ToR
> > switch.
> > > There's a much longer discussion of this scenario in Section 4.1 of
> > > draft-dunbar-nvo3-overlay-mobility-issues-00.txt .
> > >
> > > In any case, I suggest using a word other than "site" in "L2 site"
> > for
> > > clarity.
> > >
> > > Following up on Ivan's comment about tagged vs. untagged network
> > > interfaces in VMs, the first paragraph in Section 3.1 of this draft
> > > has an unfortunate mixture of VLAN-IDs assigned by the
> > > network/hypervisor
> > > with VLAN-IDs used by VMs. While both are VLAN-IDs, their
> management
> > is
> > > sufficiently different that they should be discussed separately.
> In
> > > particular, I suggest splitting the last sentence of that paragraph
> > > out into a separate paragraph on VLAN-tagged traffic to/from tagged
> > > network interfaces in VMs and providing a longer discussion
> somewhere
> > > earlier that contrasts hypervisor assignment of VLAN-IDs for
> traffic
> > > to/from untagged VM network interfaces with VLAN-ID usage by tagged
> > > VM network interfaces.
> > >
> > > Thanks,
> > > --David
> > >
> > > > -----Original Message-----
> > > > From: [email protected] [mailto:[email protected]] On
> > Behalf
> > > Of Somesh
> > > > Gupta
> > > > Sent: Wednesday, August 22, 2012 7:07 AM
> > > > To: 'Yakov Rekhter'; [email protected]
> > > > Subject: Re: [nvo3] Comments on Live Migration and VLAN-IDs
> > > >
> > > > Hopefully this is not a repeat of something hashed out earlier.
> > > >
> > > > It is not clear to me why an L2-CUG would be tied to a VLAN-ID
> > > > in an nvo3 network?
> > > >
> > > > Is this document discussing the problem as is today without
> > > > the changes nvo3 would bring? If that is the case, fine.
> > > >
> > > > Thanks
> > > > Somesh
> > > > _______________________________________________
> > > > nvo3 mailing list
> > > > [email protected]
> > > > https://www.ietf.org/mailman/listinfo/nvo3
> > >
> > > _______________________________________________
> > > nvo3 mailing list
> > > [email protected]
> > > https://www.ietf.org/mailman/listinfo/nvo3
> > _______________________________________________
> > nvo3 mailing list
> > [email protected]
> > https://www.ietf.org/mailman/listinfo/nvo3
> _______________________________________________
> nvo3 mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/nvo3
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to