RE: What we disagree on RE: TE Requirements Draft-ELSP

[Shai Herzog <sherzog@hyperchip.com>]

Nabil,

>-----Original Message-----
>From: owner-te-wg@ops.ietf.org [mailto:owner-te-wg@ops.ietf.org]On
>Behalf Of Nabil Seddigh

>I suspect most network operators would rather have a more stable
>network - one where most LSPs are not continually pre-empted due
>to a large number of pre-emption priorities. Thus, the argument
>for using a single pre-emption attribute for the E-LSP with 
>multiple BW.

Here Here. In a capitalist world, more LSP means slightly more hardware (and
slightly more expensive LSR). Compare that to cheaper LSR that may
unpredictedly melt down your network...
It will make it much harder to get to 99.999% availability, a number we all
look up to.

And an analogy: It's like trying to get the cheapest brain surgent operate
on you ;-)
Good luck!

>I understand that there are various folks interested in using
>the pre-emption attributes primarily for the case where links go
>down etc.
>
>> 3) In particular on the operational front, they both require 
>> the SP to manage things at the granularity of the OA 
>> (ie monitor/compute/configure bandwidth requirements on a 
>> per OA basis). I believe this is often the dominant scalability 
>> burden.
>
>It would be a shame to limit the protocol extensions and DS-TE
>framework because of limitations in vendor equipment path 
>computation schemes. If this argument was used, OSPF would never
>have been deployed.

Yap. How about other forms of aggregation. I also fail to understand the
scalability argument. You are talking about factor of 2-3 only on a small
subset of LSPs. 

As we all know mathematically O(multiple OA) = O(Single OA). Since they
scale the same way the efficiency and scalability argument is simply false.
If on the other hand Francois can show improvement of an order of magnitud
at least I'll understand the motivation for this weird solution.

BTW: there are other, more efficient ways of dealing with this built into
MPLS already... Why don't these SP use hierarchical MPLS tunnel aggregation?
I can show several order of magnitude improvement there (x1000?).

Shai