RE: Solicit the comments on LMP Data Channel Status I-D

[Dimitri.Papadimitriou@alcatel-lucent.be]

hello julien

i think we are diverging in this discussion - imho, the questions are 

is the proposal intended to detect running data plane failures resulting 
from some control plane component e.g. signaling ? or 
mis-use/mis-configurations triggered by user/management plane/others ?

is the proposal intended to provide near real time mechanism (fast 
detection) or an off-line mechanism (diagnostic) ? what is the diameter of 
the detection/verification such as to ensure proper localisation ?

what is the expectation from the control plane and mechanism like those 
enabled by LMP (i.e. why LMP is suitable both in terms of messaging and 
processing) ?

thanks,
-d. 









"MEURIC Julien RD-CORE-LAN" <julien.meuric@orange-ftgroup.com>
Sent by: owner-ccamp@ops.ietf.org
04/04/2007 10:44
 
        To:     Dimitri PAPADIMITRIOU/BE/ALCATEL@ALCATEL, "Diego Caviglia 
\(GA/ERI\)" <diego.caviglia@ericsson.com>
        cc:     <ccamp@ops.ietf.org>
        Subject:        RE: Solicit the comments on LMP Data Channel 
Status I-D


Hi Dimitri.

I must disagree with you. Let us get out of the pure CCAMP context to come 
to the real world.

In transmission networks, we will see both control and management planes 
being used together for a long time. In transmission networks, operators 
are used to making manual operations even if equipments are control 
plane-enabled: programmed work, configuration of legacy rings where 
control plane does not exist, maintenance... More generally, one could 
always find some specific operational cases which need human intervention 
but which does not justify the development of a specific control plane 
feature.
You may try to solve this problem by saying this pool of resource is 
"purely" (provided  this means something in terms of operations) handled 
by the control plane and that other pool is not seen; however, besides the 
fact that that kind of partitionning would still be configured manually 
with possible mistakes, you would have major drawbacks in terms of 
operations and resource optimization.

What is proposed in the ID is just a mechanism relying on a control plane 
protocol to detect *data plane* discrepencies which may occur because of 
the cross-connection nature of the circuit-switched world and of the 
manual operations you will never completely get rid of.

Regards,

Julien


-----Original Message-----
From: owner-ccamp@ops.ietf.org [mailto:owner-ccamp@ops.ietf.org] On Behalf 
Of Dimitri.Papadimitriou@alcatel-lucent.be

hi diego

not sure to understand, an automated control plane (at the end the purpose 

of this WG) is used to control XC, and other related resource reservations

now you state that there could be manual operations that fails ??? i think 

this simply falls outside the scope of the WG, on the other hand if some 
TS are unproper for cross-connection i assume the CP does not even take 
them into account, as label are downstream assigned where is the issue ?

thanks,
- d.





"Diego Caviglia \(GA/ERI\)" <diego.caviglia@ericsson.com>

Hi Dimitri,
           there can an HW problem with the unit and some of its TS are 
not good or for some reason that TS have been cross-connected by NMS for 
error or may for testing at commissioning of the TNE and due to human 
error have not been deleted and so on....

BR

Diego

-----Original Message-----
From: Dimitri.Papadimitriou@alcatel-lucent.be 
[mailto:Dimitri.Papadimitriou@alcatel-lucent.be] 

diego

> I think that the reason wht timeslot 2 is not usable is out of the
> scope of the ID.

but this is kernel of the problem, the origin tells where the solution
shall be targeted 

the i-d refers to the deletion case (2.2), fine but there is cleanup 
timeout interval to free resources in rsvp

thanks,
-d.





"Diego Caviglia \(GA/ERI\)" <diego.caviglia@ericsson.com>

This was bounced.
 
D
 

> Hi Dimitri,
>            my understunding of the ID was that it tries to 'synchronize'
> the status of the timeslots that are facing the same link.
>
> Otherwise downstream NE can choose a timeslot that is not good for the
> upstream node
>
>
> NE-A                          NE-B
> Timeslots                     Timeslots
> 1-OK                          1-OK
> 2-OK                          2-KO
> 3-OK                          3-OK
> 4-OK                          4-OK
>
> So NE-A can choose Timeslot 2 because it see it good while is not good 
in
> NE-B.  I think that the reason wht timeslot 2 is not usable is out of 
the
> scope of the ID.
>
> BR
>
> Diego
>
>
>
>
>
>
> Dimitri.Papadimitriou@alcatel-lucent.be on 03/04/2007 11.57.12
>
> diego & dan
>
> "Since the error condition can arise from a variety of situations
> (including
> control plane failure and restart, management plane intervention, 
attempt
> to clean up a partitioned control plane, etc.) we believe that it would 
be
> useful to have optional extensions to LMP to detect and report the
> problem.
> Resolution of the problem remains an issue for the management plane."
>
> we have three classes of "errors" those outside scope of the CP, those
> resulting from CP operations (bug-fixing/mis-use/mis-config/etc.) and
> then those that are strictly specific to the LMP and operations
>
> only the latter shall be in-scope - all these circular dependencies we
> are introducing in this control plane are just breaking one of the main
> design principle of networks
>
> -> which are the remaining "situations" ?
>
> thanks,
> -d.
>
>
>
>
>
> "Diego Caviglia" <Diego.Caviglia@marconi.com>
> 03/04/2007 11:35
>
>
> Hi Dan,
>        as stated in Prague I think that the ID addresses a real problem.
>
> BR
>
> Diego
>
>
>
> Dan Li <danli@huawei.com> on 03/04/2007 10.47.58
>
>
> Hi all,
>
> In Prague we saw three optical vendors stating that this I-D addresses a
> real problem that they see in deployed networks. We also saw one 
operator
> expressing interest in the work.
>
> To summarise the problem that we are addressing:
>
> Under some circumstances the opposite ends of data links may get into
> mismatched states. For example, one end of the data link may be 
allocated
> and cross-connected, while the other end is available for use. This
> represents an error condition.
>
> The problem is particularly bad when the data link is a component link 
of
> a
>
> bundle, because the problem cannot be detected from the TE 
advertisements.
> Further, existing LMP mechanisms do not allow us to discover the 
problem.
>
> If left undetected and unresolved, the situation may lead to LSP setup
> failures or to misconnection of LSPs.
>
> Since the error condition can arise from a variety of situations
> (including
>
> control plane failure and restart, management plane intervention, 
attempt
> to
> clean up a partitioned control plane, etc.) we believe that it would be
> useful to have optional extensions to LMP to detect and report the
> problem.
>
> Resolution of the problem remains an issue for the management plane.
>
> We would like to hear from people who believe that this is or is not a
> real
>
> problem in the network so that we can judge whether to ask the chairs to
> make this a WG draft.
> Best regards,
>
>
> Dan Li
>
> Advanced Technology Department
> Wireline Networking Business Unit
> Huawei Technologies Co., LTD.
> Huawei Base, Bantian, Longgang,
> Shenzhen 518129 P.R.China
> Tel: +86-755-28973237
> Fax: +86-755-28972935