Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

[Eric Gray <[e-mail]>]

Zafar,

                I believe you need to take a closer look at the relevant PCE RFCs.

--
Eric

From: Zafar Ali (zali) [mailto:[e-mail]]
Sent: Monday, February 10, 2014 12:28 PM
To: Eric Gray; John E Drake; Fatai Zhang; BRUNGARD, DEBORAH A; [e-mail]
Subject: Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03
Importance: High

Eric-

My comments were very specific to draft-zhang-ccamp-route-exclusion-pathkey-00. What I am saying is:

  *   draft-zhang-ccamp-route-exclusion-pathkey-00 requires a PCE that needs to keep state of ALL LSPs it has served, as long as LSPs exists in the network.
  *   PCE Path key is 16 bit, which presents a scaling issue with the way draft-zhang-ccamp-route-exclusion-pathkey-00 is making the Path Key "stateful".
Thanks

Regards … Zafar

From: Eric Gray <[e-mail]<mailto:[e-mail]>>
Date: Monday, February 10, 2014 11:40 AM
To: zali <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, Fatai Zhang <[e-mail]<mailto:[e-mail]>>, "BRUNGARD, DEBORAH A" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Zafar,

                Interesting, and confusing at the same time.

                Are you saying that PCE has scalability issues because network size means
there can be more tunnels?  That would seem to be true independent of PCE use.

                Or are you saying that PCE is not scalable because it needs to retain some
state information per path?  That might be one reason to put PCE functionality on a
server, rather than a router.  And it is far better that this state information is kept to
the PCE, rather than being spread around (like goo) to path ingress points.

                Keeping Path Key information at ingress points is far less of a burden on the
ingress devices than keeping the corresponding path information there.

--
Eric

From: Zafar Ali (zali) [mailto:[e-mail]]
Sent: Monday, February 10, 2014 11:32 AM
To: Eric Gray; John E Drake; Fatai Zhang; BRUNGARD, DEBORAH A; [e-mail]<mailto:[e-mail]>
Subject: Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03
Importance: High

Eric-

I was not trying to assert any requirements. I was pointing out that path key based exclusion solution has an scaling problem, which is more severe when you have 1 PCE per domain. Please note that:

1. Tunnel ID is 16 bits (any node in the network may signal 64K tunnels).
2. Path Key is also 16 bits [a given (path stateful) PCE can only keep 64K Path states].

Now as you mentioned a typical network would have 3-5 PCEs, this becomes a scaling issue for the network (e.g., 3*64 K paths for the entire network!).

Thanks

Regards … Zafar

From: Eric Gray <[e-mail]<mailto:[e-mail]>>
Date: Monday, February 10, 2014 10:50 AM
To: zali <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, Fatai Zhang <[e-mail]<mailto:[e-mail]>>, "BRUNGARD, DEBORAH A" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Zafar,

                I think you are making a pretty strong assertion in stating that "recent requirements are
to deploy a single PCE"…

                I would imagine there are many cases where multiple PCEs are used exactly as implied in
using the Path Key - in order to hide details in one routing domain from other routing domains
over which a path is to be established.

                The point I think John was making is that - where it is desirable to protect routing info
about paths - the solution you are suggesting breaks the paradigm.  This is effectively an issue
the minute there is more than one PCE, crossing multiple routing domains.

                One of the reasons for hiding internal information in one routing domain from other
routing domains is about scalability.

                It becomes a scaling issue rather quickly as well, or we wouldn't bother creating routing
domains in real networks.  There is no need to hundreds of PCEs; 5 could probably be enough,
maybe even 3.

--
Eric

From: CCAMP [mailto:[e-mail]] On Behalf Of Zafar Ali (zali)
Sent: Friday, February 07, 2014 11:47 AM
To: John E Drake; Fatai Zhang; BRUNGARD, DEBORAH A; [e-mail]<mailto:[e-mail]>
Subject: Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Hi John-

Only if you deploy a 100-s of PCE (PCE node ID-es) so scaling really becomes a function of number of PCE in the network (this does not fly). Also fact remains the  draft-zhang-ccamp-route-exclusion-pathkey-00<http://tools.ietf.org/html/draft-zhang-ccamp-route-exclusion-pathkey-00> requires a PCE that needs to keep state of ALL LSPs it has served (as long as they exists in the network). Most of the recent requirements are to deploy a single PCE, which mean the NETWORK can only serve 64K LSPs (PCE Path key is 16 bit), which also does not fly, either.

Thanks

Regards … Zafar

From: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Date: Friday, February 7, 2014 10:51 AM
To: zali <[e-mail]<mailto:[e-mail]>>, Fatai Zhang <[e-mail]<mailto:[e-mail]>>, "BRUNGARD, DEBORAH A" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: RE: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Zafar,

The conclusion of the long discussion was that your claims, repeated in your email below, regarding the path key draft were simply incorrect.

Yours Irrespectively,

John

From: CCAMP [mailto:[e-mail]] On Behalf Of Zafar Ali (zali)
Sent: Friday, February 07, 2014 7:05 AM
To: Fatai Zhang; BRUNGARD, DEBORAH A; [e-mail]<mailto:[e-mail]>
Subject: Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Hi Fatai, John-

We had a long discussion and agreement that draft-zhang-ccamp-route-exclusion-pathkey-00<http://tools.ietf.org/html/draft-zhang-ccamp-route-exclusion-pathkey-00> requires a PCE that needs to keep state of ALL LSPs it has served (as long as they exists in the network). There is also a scaling issue as it limits number of LSPs PCE can serve. So calling it is "simple" is NOT a correct statement. You solution requires a path-stateful PCE.

Also to note from last IETF meeting minutes: http://www.ietf.org/proceedings/88/minutes/minutes-88-ccamp
Lou Berger: "PCE: operating without the PCE is still in the scope of this WG." Our solution is without the use of PCE (signaling based).

Thanks

Regards … Zafar

From: Fatai Zhang <[e-mail]<mailto:[e-mail]>>
Date: Friday, February 7, 2014 1:50 AM
To: "BRUNGARD, DEBORAH A" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Re: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

Hi Deborah and all，

I think you might have noticed that there is another draft (http://tools.ietf.org/html/draft-zhang-ccamp-route-exclusion-pathkey-00), which describe a simple solution to address the same requirement as draft-ietf-ccamp-lsp-diversity and there were lots of discussion on the both drafts in the WG list.

In my understanding from the WG discussion, I think many people prefer draft-zhang because it is more straightforward and simple.

The authors of draft-zhang tried to contact the authors of draft-ietf-ccamp-lsp-diversity to discuss how to move forward these drafts, but nothing happened.

Therefore, the guidance from the WG chairs on how to move forward the both drafts (e.g, pick up one of them to go or any other suggestions) would be appreciated.




Best Regards

Fatai

From: CCAMP [mailto:[e-mail]] On Behalf Of BRUNGARD, DEBORAH A
Sent: Friday, February 07, 2014 4:06 AM
To: [e-mail]<mailto:[e-mail]>
Subject: [CCAMP] WG Last Call on draft-ietf-ccamp-lsp-diversity-03

All,

This starts a two-week working group last call on draft-ietf-ccamp-lsp-diversity-03.

This working group last call ends Feb. 20th. Please send your comments to the CCAMP mailing list.

Deborah and Lou