[dc] OVF "control plane" - Not a good idea

[<[e-mail]>]

Ashish,

Unfortunately, this is digging in the "wrong place" because it recreates the
problem that OVF was designed to solve.  OVF is intended to be a self-contained
packaging and distribution format that contains everything needed to instantiate
one or more VMs.  As such, OVF can be moved by all of the protocols noted below,
plus a variety of other means, such as sneaker-net.

If OVF is insufficient for the portability use case, then I suggest going to DMTF
to work on adding what's missing instead of inventing a "control plane" that is
at odds with OVF's design intent.

Thanks,
--David

> -----Original Message-----
> From: Ashish Dalela (adalela) [mailto:[e-mail]]
> Sent: Thursday, January 19, 2012 11:10 PM
> To: Black, David
> Cc: [e-mail]
> Subject: RE: [dc] draft-khasnabish-vmmi-problems-00.txt
> 
> 
> >> Was that supposed to be a serious question?
> 
> Yes, it is a serious question, because VM mobility goes beyond the VM.
> 
> >> If it was, I suggest FTP or NFS, both of which are already used to
> move VM
> >> images in practice, and are already specified in RFCs ;-).  OVF is
> fundamentally
> >> a VM image format.
> 
> That's one approach. Another approach is to use a SOAP/REST APIs. Yet
> another one is to define a cloud control plane, that does more than just
> move VMs. E.g. when you move a VM, you have to move the firewall rules,
> the VLAN association, the bandwidth, VRF configuration, GRE tunnel
> configuration, etc.
> 
> Thanks, Ashish
> 
> 
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]]
> Sent: Friday, January 20, 2012 8:18 AM
> To: Ashish Dalela (adalela)
> Cc: [e-mail]
> Subject: RE: [dc] draft-khasnabish-vmmi-problems-00.txt
> 
> > - Do we need a "control plane" to transfer OVF specification from
> point
> > A to B - the portability problem?
> 
> Was that supposed to be a serious question?
> 
> If it was, I suggest FTP or NFS, both of which are already used to move
> VM
> images in practice, and are already specified in RFCs ;-).  OVF is
> fundamentally
> a VM image format.
> 
> Thanks,
> --David
> ----------------------------------------------------
> David L. Black, Distinguished Engineer
> EMC Corporation, 176 South St., Hopkinton, MA  01748
> [phone]             FAX: [phone]
> [e-mail]        Mobile: [phone]
> ----------------------------------------------------
> ________________________________________
> From: [e-mail] [[e-mail]] On Behalf Of Ashish
> Dalela (adalela) [[e-mail]]
> Sent: Thursday, January 19, 2012 8:20 PM
> To: Thomas Narten; Steven Blake
> Cc: [e-mail]
> Subject: Re: [dc] draft-khasnabish-vmmi-problems-00.txt
> 
> I think it is fair to say that there is a difference between mobility
> and portability. Mobility is live migration, but portability is
> specifying a VM's properties, delete in one location and create in
> another. The new location can be another hypervisor. In many cases, you
> don't need mobility, just portability. E.g. if you have a disaster
> recovery situation, then you aren't going to get mobility anyway.
> 
> DMTF has specified a standard called OVF (Open Virtualization Format)
> that addresses the "description" of the VM. This format is supported by
> various hypervisor vendors. So, some level of VM migration
> standardization has already happened (albeit portability and not
> mobility).
> 
> The questions are:
> 
> - Do we need a "control plane" to transfer VM state from point A to B -
> the mobility problem?
> - Do we need a "control plane" to transfer OVF specification from point
> A to B - the portability problem?
> 
> The problem is relevant in the inter-datacenter, public-private, or
> inter-cloud spaces, where there will be more than one hypervisor
> controller by definition. Are we hitting the live migration issue today?
> Maybe not. Is it conceivable that we will hit this issue? I think so.
> 
> However, the question has to be asked to the provider/operators and not
> to the vendors.
> 
> Thanks, Ashish
> 
> 
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
> Thomas Narten
> Sent: Thursday, January 19, 2012 11:17 PM
> To: Steven Blake
> Cc: [e-mail]
> Subject: Re: [dc] draft-khasnabish-vmmi-problems-00.txt
> 
> Steven,
> 
> > Several system vendors (myself included) stood up in Taipei and said
> > "one encapsulation, please".  If IETF can facilitate industry
> > convergence on a small set of NVO3 encapsulations (preferably one),
> that
> > would be a big win for Ethernet switch vendors.
> 
> I agree completely.
> 
> But my questions were asking about the apparent lack of  interest from
> operators/implementers/market players regarding Bhumip's draft and the
> apparent desire to have some sort of standards work related to the
> general VM migration problem.
> 
> Is there such interest?
> 
> Thomas
> 
> _______________________________________________
> dc mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dc
> _______________________________________________
> dc mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dc