[Teas] Re: [Green] Power Aware Traffic Engineering

[Robert Raszuk <[e-mail]>]

Hi Colby,

I am observing the fact that if you have fair amount of non accounted
multicast traffic which replicates dynamically in the core of the network
based on the joins + if you have unaccounted significant chunks of unicast
TE traffic then doing any serious TE will be hard if not impossible ...
especially in pure distributed fashion.

Nothing more - nothing less.

Sure I have no shadow of a doubt that you can do a nice demo and even nicer
slides on the benefits of shutting down a line card resulting in money
savings.

PS. rfc4875 would be for accounted m-cast traffic which is not what I am
bringing here as an observation.

Cheers,
R.


On Thu, Mar 26, 2026 at 10:38 PM Barth, Colby <[e-mail]>
wrote:

> Hi Robert,
>
> Can you clarify your question or comments regarding multicast?  You seem
> to have meandered from multicast to ‘dark traffic’.
>
> As I’m sure you are aware, the procedures defined in
> https://datatracker.ietf.org/doc/html/rfc4875 and, as Tony states,
> existing TE mechanisms for resource tracking, are more than sufficient.
>  e.g. When computing paths for P2MP tunnels, the ISIS power-groups we have
> defined would be accounted for.
>
> Thanks,
>
> -- Colby
>
> *From: *Tony Li <[e-mail]> on behalf of Tony Li <
> [e-mail]>
> *Date: *Thursday, March 26, 2026 at 4:51 PM
> *To: *Robert Raszuk <[e-mail]>
> *Cc: *[e-mail] <
> [e-mail]>,
> [e-mail] <
> [e-mail]>, TEAS WG ([e-mail]) <
> [e-mail]>
> *Subject: *Re: [Green] Power Aware Traffic Engineering
>
>
> [Moving from GREEN to TEAS, GREEN to bcc]
>
>
> Hi Robert,
>
> If you want to continue disucssing this, I’m moving this conversation to
> the TEAS mailing list.  That would be more appropriate than GREEN.
>
> > > We have capacity information from our existing TE information, as well
> as existing reservations on the links.
> >
> > None of this accounts for multicast since day one.
>
>
> We are not trying to solve all problems.
>
>
> > Plus all of this requires 100% TE to account for all unicast in the
> current ISIS extensions.
>
>
> Strictly 100% is not required.  Obviously, if there is a significant
> amount of dark (i.e., non-TE) traffic, then you might get some surprising
> results.
>
>
> > Essentially today's TE is very cool to take chunk of traffic and move it
> to some other links hoping that it will not congest them.
>
>
> It’s more than hope, there is actual data.
>
>
> > But to do anything more I am of the strong opinion that current TE
> information is not enough.
>
>
> You are welcome to your opinion.
>
>
> > Especially in fully distributed computation model drying bidirectional
> links in concert from both sides is going to be fun to watch !
>
>
> I agree.  It is fun to watch and it works well.
>
> Regards,
> Tony
>
>