Re: [Lsr] LSR Flooding Reduction Drafts - Moving Forward

[Jeff Tantsura <[e-mail]>]

Les,

 

Not going to repeat Tony P. points, however I don’t think anyone said – requirement document should be a gating factor,  personally, I’d do it same way we did in RTGWG – to have a unbiased reference point others can reference to. Development of such document should go in parallel with other work and be a wg/design team effort. 

 

Hope this clarifies.

 

Cheers,

Jeff

 

From: "Les Ginsberg (ginsberg)" <[e-mail]>
Date: Wednesday, August 22, 2018 at 13:10
To: Tony Przygienda <[e-mail]>
Cc: Jeff Tantsura <[e-mail]>, Tony Li <[e-mail]>, "[e-mail]" <[e-mail]>, "Acee Lindem (acee)" <[e-mail]>
Subject: RE: [Lsr] LSR Flooding Reduction Drafts - Moving Forward

 

Tony –

 

The points you make below are good and need to be considered.

 

But let’s examine what work we have that we could do now.

 

1)draft-li-dynamic-flooding

 

This defines modest protocol extensions to support use of flooding reduction algorithms in  any type of topology. Note that it does not (not yet anyway…) specify any algorithms.

Adopting this draft would allow us to define the extensions necessary to enable the use of alternate algorithms and to get immediate benefits from the use of some well known algorithms. Since it does not preclude (and is a necessary infra for)  the use of any algorithm, and it allows support for many algorithms (NOT simultaneously though J ) it does not prevent any future algorithmic solutions from being used.

 

In parallel with this, discussion/research of points such as you mention below can be done, but I do not see why we should not proceed with this work immediately. 

 

2)Flooding reduction drafts specific to Clos topologies

 

There are multiple candidates – and I am obviously biased since I have a horse in the race (draft-shen-isis-spine-leaf-ext),  but the point is we have proposals that have practical benefits and have received support. So long as they are compatible with draft-li-dynamic-flooding I again do not see why these cannot move forward now.

 

Further research/discussion is a fine thing, but that should not prevent us from realizing real benefits now – especially since we can do so in a way that is future-proofed.

 

   Les

 

 

 

From: Tony Przygienda <[e-mail]> 
Sent: Wednesday, August 22, 2018 11:59 AM
To: Les Ginsberg (ginsberg) <[e-mail]>
Cc: Jeff Tantsura <[e-mail]>; Tony Li <[e-mail]>; [e-mail]; Acee Lindem (acee) <[e-mail]>
Subject: Re: [Lsr] LSR Flooding Reduction Drafts - Moving Forward

 

I do think it is a good idea in a sense to somehow outline WHAT problem is being solved via some write-down or mind-melt 

 

a) I hope it's captured in the meeting notes but otherwise running the danger of repeating myself, the problem splits along the line of "directed graphs" (basically lattices) which DC topologies are today and generic graphs. In first case problem can be solved quite well (Pascal's idea based loosely on MANET in RIFT that could be stretched to flat flooding as well), in 2nd it's much harder. 

b) Beside pure reduction, aspects like redundancy of the resulting mesh(es), minimal-cut properties and load balancing aspects emerge from practical pursuit of the problem (let's not even mention the dynamic re-convergence problems no matter whether some centralized instance floods or async distributed algorithm is run). Hence the scope or scopes of what needs being done seems prudent.

c) ultimately, other things like link properties and resulting meshes play a big role (MANET). In sparse networks we lived quite well without reduction but MANET/DSR had to deal with it already AFAIR & IP fabric seem to cause a different variation of the limitation to rear its head

 

2c

 

--- tony 

 

On Wed, Aug 22, 2018 at 11:04 AM Les Ginsberg (ginsberg) <ginsberg=[e-mail]> wrote:

In the discussions which led to the creation of LSVR and RIFT WGs, considerable interest was expressed in working on enhancements to existing Link State protocols. You can peruse the dcrouting mailing list archives.

 

https://mailarchive.ietf.org/arch/browse/dcrouting/

 

It is rather befuddling to me that the IETF seems to have decided to move forward on two new protocols (no objection from me) but seems to feel there is insufficient reason to move forward on proposals to extend existing IGPs.

I think the suggestion that we need to write (yet another)  requirements document before doing so is a recipe for delay – not for progress.

 

Multiple drafts have been presented over the course of the past two years and discussed on the list as well.

In the case of two of the drafts:

 

draft-shen-isis-spine-leaf-ext

draft-li-dynamic-flooding

 

WG adoption was requested in Montreal..

 

Please explain why we cannot proceed with “business as usual” as regards these drafts.

 

 

   Les

 

 

From: Lsr <[e-mail]> On Behalf Of Jeff Tantsura
Sent: Wednesday, August 22, 2018 9:43 AM
To: Tony Li <[e-mail]>
Cc: [e-mail]; Acee Lindem (acee) <acee=[e-mail]>
Subject: Re: [Lsr] LSR Flooding Reduction Drafts - Moving Forward

 

+1 Tony

 

We could start with a document, similar to dc-routing requirements one we did in RTGWG before chartering RIFT and LSVR.

Would help to disambiguate requirements from claims and have apple to apple comparison.

Doing it on github was a good experience.

 

Regards,

Jeff


On Aug 22, 2018, at 09:27, Tony Li <[e-mail]> wrote:

 

 

At IETF 102, there was no dearth of flooding reduction proposals.  In fact, we have so many proposals that there wasn’t agree as how to move forward and we agreed to discuss on the list. This Email is to initiate that discussion (which I intend to participate in but as a WG member). 

 

 

Hi Acee,

 

Perhaps a useful starting point of the discussion is to talk about requirements.  There seem to many different perceptions..

 

Regards,

Tony

 

 

_______________________________________________
Lsr mailing list
[e-mail]
https://www..ietf.org/mailman/listinfo/lsr

_______________________________________________
Lsr mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/lsr