Re: [Idr] [Can] Proposed CAN WG charter for discussion

[Robert Raszuk <[e-mail]>]

Hi Joel,

> it may be scope bounded BGP

Mind refreshing my failing memory, what is "scope bounded BGP" ?

I am aware of at least three attempts to propose it - well two failed.
Third is still waiting for any wider deployment. Perhaps I missed something
new :)

And sure I am not saying injection of host routes is needed into underlay
.. I was just describing the case where any tenant would be willing to
enter CAN augmented public cloud resources from any ISP attachment port.

Thx,
R.

On Mon, Jan 30, 2023 at 5:45 PM Joel Halpern <[e-mail]> wrote:

> It is not obvious that you need to inject the anycast address for the
> network service steering into the underlay routing.  You may.  And for some
> deployments that would be IGP, and others it may be scope bounded BGP.  But
> I can also envision solutions (which is not what we are discussing) which
> do not require such injection.  So it is not obvious that we need the
> restriction.
>
> Yours,
>
> Joel
> On 1/30/2023 3:40 AM, Robert Raszuk wrote:
>
> Hi Joel,
>
> I am not sure what is not obvious here for you (and Dirk) ?
>
> I think we all agree that the principle of operation of network centric
> CAN is based on anycast address per given compute service.
>
> Functionally you need /32 or /128.
>
> But the Internet rightfully is not going to carry anything more specific
> then /24 or /48 (maybe /64) respectively.
>
> So now to realize 10 services you must waste 10 min routable blocks.
>
> This also goes against hyperscalers injecting just /16 or /18 into the
> global Internet.
>
> So my comment in respect to applicability was not as much about dynamic
> load to the control plane introduced as it was just an observation on what
> subject solution requires vs current Internet operational rules.
>
> To summarize I do not think that bending or breaking those rules would be
> subject to analysis of this WG.
>
> Kind regards,
> R.
>
>
>
> On Mon, Jan 30, 2023 at 12:45 AM Joel Halpern <[e-mail]> wrote:
>
>> My inclination would be to leave the analysis of the public cloud
>> applicability to the working group.  Whether it is or is not applicable is
>> not obvious.
>>
>> Yours,
>>
>> Joel
>> On 1/29/2023 6:40 PM, Linda Dunbar wrote:
>>
>> Robert,
>>
>>
>>
>> You said:
>>
>> *“Also note that while you could perhaps make option 1 work in your (say
>> 5G) network for your service it does not sound like it would be applicable
>> to access public clouds compute cluster based on the actual load in the
>> same way over  the Internet.” *
>>
>>
>>
>> I think that is a very good point. Making CAN applicable to access public
>> clouds might not be wise. Maybe we should add this point to the CAN scope?
>>
>>
>>
>> My two cents,
>>
>> Linda
>>
>>
>>
>>
>
>
>