Re: [irs-discuss] 答复: 答复: IRS comments

[Thomas Nadeau <[e-mail]>]

On Aug 17, 2012:3:21 PM, at 3:21 PM, "Susan Hares" <[e-mail]> wrote:

> Tom and Mach: 
> 
> IMHO - both cases are seen in today's internet. 
> 
> +1 to Tom - Current cases such as VPNs (monitoring & interface adding) - is a current case where direct interaction may be nice
> 
> +1 to Mach - Operational cases exist where Applications orchestrate (Applications choose DC with open VM systems) and then talk to the IRS Client who commissions individual node IRS agents to perform tasks.
> 
> IMHO the IRS system is a good tool to get interact with routing system. Unlike the generic SNMP hammer, it may be a Swiss army knife tool with multiple interface methods into the system. 

	That is precisely one of its big (potential) advantages. 

	--Tom

> 
> What do you think?
> Sue  
> 
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of Thomas Nadeau
> Sent: Friday, August 17, 2012 10:09 AM
> To: Mach Chen
> Cc: [e-mail]; UTTARO, JAMES; Dutta, Pranjal K (Pranjal); Shah, Himanshu; Alia Atlas
> Subject: Re: [irs-discuss] 答复: 答复: IRS comments
> 
> 
> On Aug 16, 2012, at 10:38 PM, Mach Chen <[e-mail]> wrote:
> 
>> Hi Alia,
>> 
>> It's great to define uniform interfaces to the applications, but I am not sure whether the IRS interfaces could be directly used by the applications, in the IRS framework, the applications talk to the IRS agents, IMHO, the IRS agent (it may be the controller or orchestrator role from the SDN point of view, and the IRS interfaces are the southbound interfaces ) should shield the variety of the IRS interfaces.  
> 
> The idea is definitely to allow direct use of the interface of that is operationally desired.
> 
> Tom 
> 
>> And I agree that we should currently focus on the use cases and then define a clear scope. Here is a rough SDN related use case draft (http://tools.ietf.org/html/draft-mm-sdn-vc-vn-on-demand-use-case-01), I am not sure this is related to IRS. 
>> 
>> Best regards,
>> Mach
>> 
>>> -----邮件原件-----
>>> 发件人: Alia Atlas [mailto:[e-mail]]
>>> 发送时间: 2012年8月17日 10:13
>>> 收件人: Mach Chen
>>> 抄送: Shah, Himanshu; UTTARO, JAMES; Dutta, Pranjal K (Pranjal); 
>>> [e-mail]
>>> 主题: Re: 答复: [irs-discuss] IRS comments
>>> 
>>> Hi Mach,
>>> 
>>> Right - that is one reason why I'm not sure that a CSPF computation 
>>> piece is needed for the RSVP sub-interface.  On the other side, 
>>> there's benefit to not requiring an application to understand too 
>>> many different protocols.
>>> 
>>> We need to discuss what scope makes sense for the sub-interfaces.  I 
>>> think the use-cases will drive this.
>>> 
>>> Alia
>>> 
>>> On Thu, Aug 16, 2012 at 10:06 PM, Mach Chen <[e-mail]>
>>> wrote:
>>>> Hi Alia,
>>>> 
>>>> I am a little confused, the said RSVP sub-interface (on-demand CSPF 
>>>> interface)
>>> seems what exactly PCEP does today. My understanding of IRS is that 
>>> it may be complementary on topology discovery.
>>>> 
>>>> Best regards,
>>>> Mach
>>>> 
>>>>> -----邮件原件-----
>>>>> 发件人: [e-mail] 
>>>>> [mailto:[e-mail]]
>>> 代
>>>>> 表 Alia Atlas
>>>>> 发送时间: 2012年8月17日 9:45
>>>>> 收件人: Shah, Himanshu
>>>>> 抄送: UTTARO, JAMES; Dutta, Pranjal K (Pranjal); [e-mail]
>>>>> 主题: Re: [irs-discuss] IRS comments
>>>>> 
>>>>> I'm not entirely clear on the model or question you are asking.  An 
>>>>> application could learn topology via IRS to the appropriate IRS 
>>>>> agents, do CSPF computations, and support a signaling/CSPF-like 
>>>>> sub-interface (if we define such for IRS) as an IRS agent.  Another 
>>>>> application could query the first for the CSPF results and then use 
>>>>> IRS to trigger signaling on the appropriate device.
>>>>> 
>>>>> I'm not sure if it makes sense to have an IRS sub-interface that 
>>>>> requests CSPF computation, but it might come as part of an RSVP 
>>>>> sub-interface to trigger signaling.  We'll have to see about 
>>>>> use-cases.
>>>>> 
>>>>> I don't see a strong motivation to duplicate existing work.
>>>>> 
>>>>> Alia
>>>>> 
>>>>> On Thu, Aug 16, 2012 at 2:46 PM, Shah, Himanshu <[e-mail]>
>>> wrote:
>>>>>> I was wondering if PCE and IRS could have overlapping functionality.
>>>>>> 
>>>>>> Meaning, can an IRS interface user at both ends (node and 
>>>>>> controller) be a
>>>>> non-PCE entity and perform
>>>>>> the same service as what PCE does?
>>>>>> 
>>>>>> It seems like you are saying that could work co-operatively...
>>>>>> 
>>>>>> Thanks,
>>>>>> himanshu
>>>>> 
>>>>> ==snip===
>>>>> _______________________________________________
>>>>> irs-discuss mailing list
>>>>> [e-mail]
>>>>> https://www.ietf.org/mailman/listinfo/irs-discuss
>> _______________________________________________
>> irs-discuss mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/irs-discuss
> _______________________________________________
> irs-discuss mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/irs-discuss
> 
>