Re: [irs-discuss] An idea ... MTR + IRS

Tina,

IRS makes no statement about orchestrators; I'm pretty sure the word
doesn't appear in either draft at all.  Certainly an orchestrator is a
type of application that could use IRS.

IRS includes routing state, topology, and the ability to get
measurements and dynamic events.  All may be needed for a meaningful
feedback loop run by an application.

Alia

On Wed, Aug 8, 2012 at 1:24 AM, Tina TSOU <Tina.Tsou.Zouting@huawei.com> wrote:
> Alia,
> It is just to clarification.
> Just to confirm that draft-atlas and draft-ward state that IRS positions below the northboudapi in an orchestrator, and IRS includes both routing state and topo export. Is it correct understanding?
> Thank you.
>
> Tina
>
>
>> -----Original Message-----
>> From: Alia Atlas [mailto:akatlas@gmail.com]
>> Sent: Tuesday, August 07, 2012 9:38 PM
>> To: Tina TSOU
>> Cc: Edward Crabbe; robert@raszuk.net; irs-discuss@ietf.org
>> Subject: Re: [irs-discuss] An idea ... MTR + IRS
>>
>> Tina,
>>
>> Nitin made a single comment on this list that you are reading too much
>> into.
>>
>> I have written the two base drafts for IRS and am working on driving
>> the definition of the work that is proposed.  The presentation that I
>> gave in Routing Area Open Meeting CLEARLY describes topology as not
>> only being in the proposed scope, but one of the more urgent items.
>>
>> Why are you NOT listening?
>>
>> Perhaps rereading the problem-statement and framework drafts would
>> help clarify your mental model?
>>
>> IRS does not talk about a northboundapi - there are applications that
>> can use IRS.  How those applications communicate with other
>> applications is NOT in the proposed scope.
>>
>> Alia
>>
>> On Wed, Aug 8, 2012 at 12:07 AM, Tina TSOU
>> <Tina.Tsou.Zouting@huawei.com> wrote:
>> > So, r u updating Nitin's statement as following?
>> >
>> > Top layer: Northboundapi
>> >
>> > Middle layer: IRS
>> >
>> > Bottom layer: IRS (topo export)
>> >
>> >
>> > Tina
>> >
>> > On Aug 7, 2012, at 9:01 PM, "Alia Atlas" <akatlas@gmail.com> wrote:
>> >
>> > The requirement for topology export is DEFINITELY part of IRS.  It is
>> > a crucial piece for a meaningful feedback loop.   Clearly there are
>> > existing technology pieces that may have a role to play as well.
>> >
>> > Alia
>> >
>> > On Tue, Aug 7, 2012 at 10:40 PM, Tina TSOU
>> <Tina.Tsou.Zouting@huawei.com>
>> > wrote:
>> >
>> > Oh, I meant in Nitin's description in the mailing list, there are 3
>> > sub-layers of a orchestrator.
>> >
>> > Top layer: Northboundapi
>> >
>> > Middle layer: IRS
>> >
>> > Bottom layer: topo export
>> >
>> >
>> > Therefore, the topo export is not part of IRS. It is another layer
>> which
>> > sits below IRS.
>> >
>> >
>> > Tina
>> >
>> >
>> > On Aug 7, 2012, at 5:30 PM, "Alia Atlas" <akatlas@gmail.com> wrote:
>> >
>> >
>> > Topology export is definitely under the umbrella of IRS - and we are
>> >
>> > actively starting to think about the associated requirements and
>> >
>> > use-cases.   Feel free to contribute - on the list or towards drafts!
>> >
>> >
>> > Alia
>> >
>> >
>> > On Tue, Aug 7, 2012 at 6:23 PM, Tina TSOU
>> <Tina.Tsou.Zouting@huawei.com>
>> > wrote:
>> >
>> > Nitin said topo information export is on the sub-layer under IRS.
>> >
>> >
>> >
>> >
>> > Tina
>> >
>> >
>> >
>> >
>> > From: irs-discuss-bounces@ietf.org [mailto:irs-discuss-
>> bounces@ietf.org] On
>> >
>> > Behalf Of Edward Crabbe
>> >
>> > Sent: Tuesday, August 07, 2012 3:20 PM
>> >
>> > To: robert@raszuk.net
>> >
>> > Cc: irs-discuss@ietf.org
>> >
>> >
>> >
>> > Subject: Re: [irs-discuss] An idea ... MTR + IRS
>> >
>> >
>> >
>> >
>> > Robert,
>> >
>> >
>> >
>> >
>> > If topo information export is in scope (which I believe it is) and
>> PBR route
>> >
>> > injection with nh recursion to rib (and thus connected routes) is in
>> scope
>> >
>> > (which I'm quite sure it is) then yes, this is in scope.
>> >
>> >
>> >
>> >
>> > Although I'm not sure what it has to do with OF /OF controllers? ;P
>> >
>> >
>> > On Thu, Aug 2, 2012 at 5:02 PM, Robert Raszuk <robert@raszuk.net>
>> wrote:
>> >
>> >
>> > Hi,
>> >
>> >
>> > This morning Scott mentioned that he would like to use IRS to shut
>> down all
>> >
>> > protocols and just be able to write to RIB. Now James said that he
>> would
>> >
>> > like to get a network topology as "every OpenFlow controller requires
>> this"
>> >
>> >
>> > Both connected together resulted in an idea of using multi-topology-
>> routing
>> >
>> > where your base topology discovers physical link connectivity graph
>> while
>> >
>> > other topologies could be programmed by external entities example: OF
>> >
>> > controllers or any other external to routers network intelligence
>> oracles to
>> >
>> > deliver actual services ?
>> >
>> >
>> > Would that be in scope of IRS effort ? If so what would be the
>> proposed
>> >
>> > "write to RIB" set of protocols ? Would you support OF 1.3 even if
>> one would
>> >
>> > be happy to lock such topologies only to software/programmable
>> switching
>> >
>> > paths ?
>> >
>> >
>> > Best rgs,
>> >
>> > R.
>> >
>> > _______________________________________________
>> >
>> > irs-discuss mailing list
>> >
>> > irs-discuss@ietf.org
>> >
>> > https://www.ietf.org/mailman/listinfo/irs-discuss
>> >
>> >
>> >
>> >
>> >
>> > _______________________________________________
>> >
>> > irs-discuss mailing list
>> >
>> > irs-discuss@ietf.org
>> >
>> > https://www.ietf.org/mailman/listinfo/irs-discuss
>> >
>> >