Re: [L3sm] Comments on draft-ietf-l3sm-l3vpn-service-model

[Qin Wu <[e-mail]>]

Not from operators, my take on this is modularization approach is really important for service automation, service agility.
Based on the L3SM prototype demonstrated in Prague Bits-N-Bytes, modular approach enables faster service creation, modification, e.g., we can use site template to create site more quickly,
I am happy to see the new version of draft-ietf-l3sm-l3vpn-service-model get in line with this idea.
In addition, modularization approach can help us to develop other new service model more quickly by reusing common components defined in the existing service model.

-Qin
发件人: L3sm [mailto:[e-mail]] 代表 Scharf, Michael (Michael)
发送时间: 2015年7月23日 18:40
收件人: Rob Shakir; 'Kireeti Kompella'; [e-mail]; [e-mail]; [e-mail]
主题: Re: [L3sm] Comments on draft-ietf-l3sm-l3vpn-service-model

Hi Rob,

My understanding is that the abstract L3SM model would be offered by a network management system software. Being in R&D, I mean the software “implementation” of the L3SM model.

I fully agree that deploying a corresponding NMS solution is not trivial, in particular if the network operator’s services change. But I don’t really understand why a more modular abstract service model in L3SM would simplify these challenges. But for these type of questions we would really appreciate input from network service providers.

Thanks

Michael


From: Rob Shakir [mailto:[e-mail]]
Sent: Thursday, July 23, 2015 12:19 PM
To: 'Kireeti Kompella'; Scharf, Michael (Michael); [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [L3sm] Comments on draft-ietf-l3sm-l3vpn-service-model





On 23 July 2015 at 11:15:26, Scharf, Michael (Michael) ([e-mail]<mailto:[e-mail]>) wrote:
> > But some early, precautionary modularisation is not going to be a
> problem so long as we don't spend too long arguing about exactly what
> to modularise.
>
> Fully agree ..., some sections Kireeti was mentioning are evident to be
> as a reusable component so we can do it now, for more complex things,
> let's think about it after the module is (almost) finished.

There is great value in having a consistent model without complex external dependencies, and we should aim at publishing that model soon. This significantly simplifies implementation.

Can you clarify which implementation you mean here, please? IMHO — there is a set of one off implementation effort to support the model in a network mgmt system - but there is the implementation of changes across a set of services offered by an operator. I expect that more changes happen within the operator’s services (new hardware, new constraints from suppliers or external network dependencies being removed, new customer requirements) than happen for the NMS needing to support different model features (from what I see today), hence, it’d be good to have a debate as to which implementations we want to simplify.

An approach that makes changes to services hard or costly to implement for operators will result in any models produced in the IETF being unusable, or not helping with the current complexities we have.

Best,

r.