Re: [irs-discuss] IRS comments

["Shah, Himanshu" <[e-mail]>]

I will let Ed clarify on what he means by "controller" .

But here is my view of "application", "controller", "use-case" etc etc,

When an XaaS processes a request it takes into consideration of availability of the CPU, memory, application data, cost of power, etc. to instantiate a VM on some host machine.
One additional resource taken into consideration is the availability and/or requirement of the network connectivity between the resources
such as VM, application data, requester, etc.

The provider of XaaS thingy is the application server.
(Centralized) Network "Controller(s)"  knows current state of network infrastructure and (to be decided how much it) controls member network nodes using IRS.
The "use case" here is to be able to dynamically create L2/L2.5/L3 connectivity with specific TE characteristics between the (perhaps geographically dispersed) resource points.

In hierarchical fashion:   Application server <- (application interface) -> Network Controller <- (IRS) -> Network Node

As we heard at IETF, IRS has tentacles in network nodes from BGP policies, all the way down to FIBs/LFIBs/ACL.

So we need use cases for which applications would require accessibility to -
BGP Policies
RIB
LSDB ( I saw an email which talks about reducing IGP to link distribution protocol and running SPF in centralized network controller)
LIB
FIB
ACL (this is perhaps obvious)
Etc etc

I broad brushed and simplified a lot here to express my view - not sure if this jives with others.

/Himanshu








From: [e-mail] [mailto:[e-mail]] On Behalf Of Olen Stokes
Sent: Wednesday, August 15, 2012 1:30 PM
To: Edward Crabbe; Alia Atlas
Cc: David Lake (dlake); [e-mail]
Subject: Re: [irs-discuss] IRS comments

Thanks.  Can you also give us what you mean by "controller".

Olen

From: Edward Crabbe [mailto:[e-mail]]<mailto:[mailto:[e-mail]]>
Sent: Wednesday, August 15, 2012 1:24 PM
To: Alia Atlas
Cc: David Lake (dlake); Olen Stokes; [e-mail]<mailto:[e-mail]>
Subject: Re: [irs-discuss] IRS comments

s/wg/pre-BOF proto-wg :P/g
On Wed, Aug 15, 2012 at 10:22 AM, Edward Crabbe <[e-mail]<mailto:[e-mail]>> wrote:
+1 Alia.  There's been a lot of confusion over this term.  Having gone a few rounds with folks on this one in other forums, I'll point out that what most people mean by application (myself included) is some set of control software (a scheduler, a path optimizer etc)  that provides instructions to the controller, which are in turn translated to the appropriate PDUs.

Having 'end user' applications request/make changes to forwarding state without an intermediate broker/aggregator acting on their behalf sounds like a recipe for disaster for operational networks, or, as is more likely, a quick hike to the protocol grave yard (followed by a long grave-side party :P) for the wg.

my 2c.

On Wed, Aug 15, 2012 at 8:48 AM, Alia Atlas <[e-mail]<mailto:[e-mail]>> wrote:
Hi David,

We do need to clarify what is meant by an application.  I would not
expect that real user-land applications would talk directly to routing
devices via IRS.  I can see that going through an intermediary.  The
IRS abstractions are unlikely to be as high-level as user-land
applications would want and the security and policy issues would get
exciting.

Clarifying what applications are more in-scope initially is part of
where use-cases will help.  Can you write up ones to
categorize/describe your thoughts?

Alia

On Wed, Aug 15, 2012 at 11:40 AM, David Lake (dlake) <[e-mail]<mailto:[e-mail]>> wrote:
> As another newbie to this, I have some questions about "application vendors."
>
> Who is the target audience here ?   That will determine what functionality and abstraction of the network we need to expose to that "application."
>
> This presently appears to be a little confused - at least in my mind.  The draft talks very much as if the application we are addressing is an OSS/BSS system, essentially provisioning from the domain owner.
>
> However, linking this to the wider goals of SDN as voiced by customers/users at the first Open Network Summit, there appears to be a desire for cross-domain and user-land application integration.
>
> At this level - as an example giving a content cache the ability to ensure delivery of an HD video to an end user - the application will not be interested in the underlying topology of the network; it will  need to know that application X can be delivered with parameters Y between reading from the content store to delivery to the user's browser.   How the stream traverses the infrastructure is immaterial.
>
> Are we intending that IRS satisfies BOTH these requirements (i.e. for ALL applications ?), or should we be more prescriptive about which application space we are addressing ?
>
> Thanks
>
> David
>
> -----Original Message-----
> From: [e-mail]<mailto:[e-mail]> [mailto:[e-mail]<mailto:[e-mail]>] On Behalf Of Alia Atlas
> Sent: Wednesday, August 15, 2012 7:23 AM
> To: Olen Stokes
> Cc: [e-mail]<mailto:[e-mail]>
> Subject: Re: [irs-discuss] IRS comments
>
> I have not specifically heard from application vendors about this.
> My current plan is that we focus on a Use-Cases draft and define within that some motivating use-cases that we agree are good first targets.  Those can drive which subset of functionality we focus on.
>
> More use-cases are, of course, quite welcome.  Posting them to the mailing list is a good first start.  Russ White is starting the general use-cases draft based on the three use-cases that he sent to the list.
>
> Alia
>
> On Wed, Aug 15, 2012 at 9:43 AM, Olen Stokes <[e-mail]<mailto:[e-mail]>> wrote:
>> Are there applications vendors out there that already have specific requirements for what this " subset of the data-models for sub-interfaces"  should be?
>>
>> Olen
>>
>> -----Original Message-----
>> From: [e-mail]<mailto:[e-mail]>
>> [mailto:[e-mail]<mailto:[e-mail]>] On Behalf Of Alia Atlas
>> Sent: Wednesday, August 15, 2012 9:08 AM
>> To: Shah, Himanshu
>> Cc: Gert Grammel; [e-mail]<mailto:[e-mail]>; Lenny Giuliano; Thomas Nadeau;
>> Alia Atlas; Scott Whyte
>> Subject: Re: [irs-discuss] IRS comments
>>
>> Hi Himanshu,
>>
>> Welcome.   I agree that IRS isn't going to spring into being fully
>> formed - I expect that we'll focus on a subset of the data-models for sub-interfaces along with an associated protocol (whether that is a new one or extending an existing one).
>>
>> Alia
>>
>> On Wed, Aug 15, 2012 at 7:41 AM, Shah, Himanshu <[e-mail]<mailto:[e-mail]>> wrote:
>>> Newbie to this discussions list and have read only a last couple of mails, so pardon the repeat if somebody has already raised the following as a concern.
>>>
>>> I realize we are early in IRS architecture definition but one thing to keep in mind is the user experience.
>>> We need to make sure that exposed interface to
>>> RIB/LFIB/FIB/IGPs/BGP/LSDBs etc etc  provide a consistent predictive action/response/events even when different implementations has varying capabilities.
>>>
>>> At the moment it seems like a wild wild west.
>>> Perhaps IRS can be defined in phases starting with a simpler, limited version..
>>>
>>> Thanks,
>>> himanshu
>>>
>>>
>>> -----Original Message-----
>>> From: [e-mail]<mailto:[e-mail]>
>>> [mailto:[e-mail]<mailto:[e-mail]>] On Behalf Of Alia Atlas
>>> Sent: Monday, August 13, 2012 8:41 AM
>>> To: Scott Whyte
>>> Cc: Thomas Nadeau; Gert Grammel; Alia Atlas; Lenny Giuliano;
>>> [e-mail]<mailto:[e-mail]>
>>> Subject: Re: [irs-discuss] IRS comments
>>>
>>> ...snip...
>>>
>>> On Fri, Aug 10, 2012 at 9:03 PM, Scott Whyte <[e-mail]<mailto:[e-mail]>> wrote:
>>>> On Fri, Aug 10, 2012 at 5:16 PM, Alia Atlas <[e-mail]<mailto:[e-mail]>> wrote:
>>>
>>>>> I do think it is important to have the RIB as an arbitration mechanism
>>>>> on the device.   Russ's suggestion that for the RIB sub-interface, the
>>>>> IRS agent might communicate logically to an IRS routing process
>>>>> gives good semantics and interactions.  Obviously, implementations
>>>>> may differ.
>>>>
>>>> As long as the arbitration mechanism is reconfigurable by the
>>>> operator to whatever precedence they want, I agree.  Its not clear
>>>> to me if various RIB implementations treat all proffered routes the
>>>> same, nor if they store the same meta-data with all protocol sources.
>>>> So there needs to be some way for the operator to leverage exposed
>>>> protocol-specific optimizations, without conflict from the other
>>>> routing processes, if they so desire.  OTOH if it can all be done
>>>> via static routes, it seems much simpler. :)
>>>
>>> Clearly the IRS sub-interface for the RIB needs to introduce/define the different precedences; my assumption is that it would be per route with a well-defined small set of meta-data.  This is part of where having good use-cases will help us understand what behavior is necessary.  The static  routes do seem like a simpler case to start with.
>>>
>>> Alia
>>> _______________________________________________
>>> irs-discuss mailing list
>>> [e-mail]<mailto:[e-mail]>
>>> https://www.ietf.org/mailman/listinfo/irs-discuss
>> _______________________________________________
>> irs-discuss mailing list
>> [e-mail]<mailto:[e-mail]>
>> https://www.ietf.org/mailman/listinfo/irs-discuss
> _______________________________________________
> irs-discuss mailing list
> [e-mail]<mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/irs-discuss
_______________________________________________
irs-discuss mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/irs-discuss