Re: [CCAMP] IETF 89 Draft Agenda Posted

[Daniele Ceccarelli <[e-mail]>]

Gert, (all)

You're right, and that's why a good planning of the network is fundamental. 
However a "fine tuning" of the network is needed, in all of those cases that were not included in the planning phase (e.g. some additional service requests that are not enough to justify another planning iteration, e.g. extremely dynamic networks where the traffic matrix highly changes in time and it is not possible to perform a good planning). 

This is why I tend to think that:
1. the perfect overlay network both implements the a priori planning of the network with potential server layer LSPs *and* the UNI with "steroids".
2. The "steroids" are useful, but too many are detrimental: I think that the UNI should be augmented with just some path computation constraints (OF , TE bounds and diversity) and the capability of collecting some TE metrics. The rest is "gilding the lily" (cit. John). If some many features are needed in an overlay network, that the a priori topological info approach is more suitable.
3. The statement ". If there is no TE needed, UNI/overlay is adequate." Is not fully correct for what said above

BR
Daniele

> -----Original Message-----
> From: CCAMP [mailto:[e-mail]] On Behalf Of Gert Grammel
> Sent: lunedì 24 febbraio 2014 20:38
> To: John E Drake; Lou Berger; Fatai Zhang; CCAMP
> Cc: Alia Atlas
> Subject: Re: [CCAMP] IETF 89 Draft Agenda Posted
> 
> Hi John,
> 
> The real difficulty of those requests starts even earlier, since the requestor
> (aka. client) doesn't have visibility about what's possible and hence needs to
> narrow-down in a request-response manner the best choice for himself.
> Imagine a requestor would request 4 connections between A and B. the
> default choice are: all co-routed or all diverse. If neither of both works,
> what's next? Asking for 3 diverse routed or joint path and then for a third
> free routed? Go for two requests with 2 path each? ...
> 
> An overlay network works well if it is reasonable that the request can be
> serviced. In a Telephony network with 1:1 connections, good statistics and
> lots of over-provisioning this can be fulfilled. *If* the network is designed in
> an A/B structure even a 1+1 protection connectivity can be reasonably
> requested. However in a client/server network that is not A/B and needs to
> cover different diversity and resource constraints that's not realistic. It is just
> like with other TE constraint parameters like Latency. If a client needs to
> dream up a latency requirement he would probably select the geographic
> distance, multiply it by speed of light and a rule-of-thumb parameter and ask
> for it. What else could he do?
> 
> UNI/Overlay is not made for traffic engineering or client controlled
> redundancy. If TE is important, a-priori topological information is required. If
> there is no TE needed, UNI/overlay is adequate.
> 
> Best
> 
> Gert
> 
> -----Original Message-----
> From: CCAMP [mailto:[e-mail]] On Behalf Of John E Drake
> Sent: 24 February 2014 17:08
> To: Lou Berger; Fatai Zhang; CCAMP
> Cc: Alia Atlas
> Subject: Re: [CCAMP] IETF 89 Draft Agenda Posted
> 
> Hi (and copying Alia and Adrian),
> 
> I've included Suurballe's original paper for background reading on the subject
> of computing N node disjoint paths through a network.  The careful reader
> will note that the N node disjoint paths need to be computed concurrently as
> otherwise it may not be possible to find disjoint paths.
> 
> The implication of this is that all of the current proposals, which attempt to
> find disjoint paths sequentially are fundamentally broken.
> 
> If we want to get N node disjoint paths across, I think we need to define a
> new service in which the ingress CE sends the ingress PE a request containing
> a destination (egress) CE, the number of disjoint paths N, and a list of its
> attachment circuits to be considered in the request.  What it receives in
> exchange is a list of [attachment circuit, path keys] which it then can include
> in a set of Path messages that it sends to the provider network.
> 
> However, this is really a service-level request rather than an LSP
> establishment request and the while the latter is clearly within CCAMP's
> charter, I think the former is more in the purview of the PCE working group
> where a stateful PCE can provide this service, if it doesn't already.
> 
> In any case, to proceed with any of the current proposals is irresponsible.
> 
> Yours Irrespectively,
> 
> John
> > -----Original Message-----
> > From: CCAMP [mailto:[e-mail]] On Behalf Of Lou Berger
> > Sent: Wednesday, February 19, 2014 7:28 AM
> > To: Fatai Zhang; CCAMP
> > Subject: Re: [CCAMP] IETF 89 Draft Agenda Posted
> >
> > Fatai,
> > 	Our intent is to have the current WG draft authors propose how they
> > think the document should move forward based on the WG LC comments.
> > Given that we are not meeting until Thursday, I'd hope that WG draft
> > authors take advantage of this and work/meet with the authors of the
> > related drafts to come up with an approach that is inclusive of the
> > issues and alternatives mentioned.
> >
> > Lou
> >
> > On 2/18/2014 10:39 PM, Fatai Zhang wrote:
> > > Hi Lou,
> > >
> > > Do you think if it is better to put presentation #4,#12,#13 close
> > > (ie., treat
> > them as a group topic like diversity route)?
> > >
> > >
> > >
> > >
> > >
> > > Best Regards
> > >
> > > Fatai
> > >
> > >
> > > -----Original Message-----
> > > From: CCAMP [mailto:[e-mail]] On Behalf Of Lou Berger
> > > Sent: Tuesday, February 18, 2014 5:58 AM
> > > To: CCAMP
> > > Subject: [CCAMP] IETF 89 Draft Agenda Posted
> > >
> > >
> > > All,
> > > 	The draft agenda for London was posted earlier today at:
> > > http://www.ietf.org/proceedings/89/agenda/agenda-89-ccamp
> > >
> > > Please let us (Daniele & chairs) know if you have comments or if we
> > > missed anything.
> > >
> > > Authors of WG documents,
> > >
> > > As usual:
> > >
> > > If presenting, please plan to review (present) any changes that have
> > > been recently made, any open discussions or issues, as well as
> > > planned next steps.
> > >
> > > If you are not presenting, please send this information to the WG
> > > mail list *1 week* prior to the WG meeting.
> > >
> > > Much thanks,
> > > Lou, Deborah (& Daniele)
> > >
> > > _______________________________________________
> > > CCAMP mailing list
> > > [e-mail]
> > > https://www.ietf.org/mailman/listinfo/ccamp
> > >
> > >
> > >
> > >
> >
> > _______________________________________________
> > CCAMP mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/ccamp
> >
> 
> 
> _______________________________________________
> CCAMP mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/ccamp