[openstack-dev] [all][massively distributed][architecture]Coordination between actions/WGs

Curtis serverascode at gmail.com
Mon Aug 29 23:27:25 UTC 2016

On Mon, Aug 29, 2016 at 2:15 PM, Joshua Harlow <harlowja at fastmail.com> wrote:
> Curtis wrote:
>> On Mon, Aug 29, 2016 at 1:27 PM, gordon chung<gord at live.ca>  wrote:
>>> just to clarify, what 'innovation' do you believe is required to enable
>>> you
>>> to build on top of OpenStack. what are the feature gaps you are
>>> proposing?
>>> let's avoid defining "the cloud" since that will give you 1000 different
>>> answers if you ask 1000 different people.*
>> One idea I hear fairly often is having a couple of hypervisors in say
>> a single store or some other customer premise, but not wanting to also
>> run an OpenStack control plane there. If we are talking about a
>> hypervisor level, not some other unknown but smaller IoTs...uh thing,
>> does that make more sense from a OpenStack + vCPE context? Or do some
>> think that is out of scope for OpenStack's mission as well?
>> Thanks,
>> Curtis.
> <snip>
> So is that like making a customer premise have the equivalent of dumb
> terminals (maybe we can call them 'semi-smart' terminals) where those things
> basically can be remote controlled (aka the VMs on them can be upgraded or
> downgraded or deleted or ...) by the corporate (or other) entity that is
> controlling those terminals?
> If that's the case, then I don't exactly call that a cloud (in my classical
> sense), but more of a software delivery (and remote-control) strategy (and
> using nova to do this for u?).
> But then I don't know all the 3 or 4 letter acronyms so who knows, I might
> be incorrect with the above assumption :-P

>From a brief look, it seems like vCPE is more along the lines of the
customer having a "thin" device on their premise and their (now
virtual) network functions, eg. firewall, live in the providers data
center over a private link created by that thin device. So having a
hypervisor on a customer premise is probably not what most telecoms
would consider vCPE [1].

But in my (limited) example, I'm not talking about managing that thin
device, I am thinking of a hypervisor or two instead in a customer
premise, or remote location, that is controlled by some (magic?)
remote nova, and yeah, would have access to glance, etc, to deploy
instances, basically as a way of avoiding running an OpenStack control
plane there. But not so much in the way of managing upgrades of the
software on that virtual machine on that hypervisor or anything, just
acting as IaaS.


[1]: Pg. 48 - http://innovation.verizon.com/content/dam/vic/PDF/Verizon_SDN-NFV_Reference_Architecture.pdf

> -Josh
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

Blog: serverascode.com

More information about the OpenStack-dev mailing list