RE: Question on LMP.

[Zafar Ali <zali@cisco.com>]

Hi Ravi, 

Please see comments in-lined. 

Thanks

Regards... Zafar 

At 03:08 PM 4/22/2002 -0400, Ravi Ravindran wrote:
> hi
> zafar,
> You have stated that LMP
> hellos should be fast enough to detect CC failure before the routing and
> the signalling protocols (belonging to a GMPLS control plane),

I am sorry for the confusion but I was NOT referring to the IGP Hellos in
GMPLS control network. Instead I was referring to the IGP Hellos that are
running from a router-to-router in the L3 network that is using the
optical LSPs (or data bearer channels in the optical network). 

> but
> then what is the action taken by LMP after that, does LMP interact with
> the L3 protocols (incase there are multiple LMP active CC's b/w two
> nodes) leading to update of the routing tables. 

No, 

> how
> does it prevent loosing IGP hellos (section 3.2)
> ?
>  

Again, please note that I am referring to the L3 adjacency that are
established from a router-to-router in the L3 network that is using the
optical LSPs. 

Recall, in the case of RSVP-TE, if the last control channel fails and
does not come back up within "RSVP restart time" or before
expiration of "cleanup timeout" interval related
to the refresh messages, data forwarding on the optical LSPs in question
can no longer be sustained. Hence, in order to avoid such data losses or
possible loss of the IGP
adjacency, one may
adapt a conservative approach in reacting to the failure of the last
control channel, e.g., triggering path/ local protection, a
make-before-break scheme, etc.  Of course, this is beyond the scope
of current discussion as well as the LMP draft.

thanks

Ravi S. Ravindran
Nortel Networks, Ottawa
 

 -----Original Message-----
From: Zafar Ali [mailto:zali@cisco.com]
Sent: Monday, April 22, 2002 12:47 PM
To: Michiel van Everdingen
Cc: Manoj Sontakke; Jonathan Lang; ccamp@ops.ietf.org
Subject: Re: Question on LMP.

Hi Michiel,

Please see some comments in-lined.

Thanks

Regards... Zafar

At 08:32 AM 4/22/2002 +0200, Michiel van Everdingen wrote:

Hello Manoj, Ravi, Jonathan, Zafar,

I'm somewhat confused on the state of this thread. As far as I

can see, the following questions are still open:

- why is control channel management needed at all ?

The control channel failure detection mechanism is required if lower-level mechanisms are NOT available to detect control channel failures. E.g., when control channel is (IP) routed and not bound with a particular interface. Furthermore, a control channel failure is an event on which applications are interested in from various prospective. E.g., we would like to distinguish between signalling channel failure and control channel failure during RSVP-TE recovery process, etc.

- why does control channel management need to be fast ?

A note on the frequency of LMP Hellos: Please note that we need to distinguish between a signaling channel failure and the control channel failure. Hence, LMP Hellos should be faster than RSVP Hellos or the mechanism used to detect signaling channel failure. Similarly, LMP Hello frequency should be greater than IGP hello frequency, so that the optical network can make "conscious" decision on the control channel failure, before having an adverse affect on the IGP adjacencies at L3.

- is control channel management fast ?

I think the LMP Hello frequency need to follow the above mentioned criteria. By fast, do you mean O(milliseconds)? If yes, I don't think LMP Hellos need to be "fast".

As stated in section 13 of the LMP draft, all LMP messages are

IP encoded. So a standard general purpose IP network should

perfectly well be capable to transport such IP encoded LMP

messages to the intended destination. Am I missing something

here ?

I am sorry but I did not understand what you meant here?

One of the advantages of using a general purpose IP network is

that there is no need for a point-to-point direct communication

channel between sender and destination. point-to-point direct

communication can become a problem, e.g. when the dataLink is a

VC-12 or a VT-1.5.

Moreover, control channel management seems to give unnecessary

overhead in case there is only one interface between two

neighbouring switches.

IMO when control channels are interface bound (i.e., failure of the interface means failure in the control channel) and L2 mechanism are available to for failure detection, we can run LMP Hellos at a lower frequency and rely on L2 control channel failure detection. But please note that this is not always the case that your control channel are bound to a particular interface (e.g., IP routed control channels), hence the need for failure detection within the scope of LMP.

<snip>
=================
Zafar Ali
Cisco Systems
(734) 276-2459
100 S Main St. #200
Ann Arbor, MI 48104.