The MPLS-OPS Archive

[Aamer Akhter <aakhter@cisco.com>]

On Oct 26, 2004, at 1:51 AM, Spice Sylvia wrote:

> Hello Aamer,
>
>> On Oct 25, 2004, at 1:06 AM, Spice Sylvia wrote:
>>
>>> Hi Aamer,
>>>
>>>
>>>  --- Aamer Akhter <aakhter@cisco.com> wrote:
>>>>
>>>> On Oct 21, 2004, at 4:29 AM, Spice Sylvia wrote:
>>>>
>>>>> Hello,
>>>>>
>>>>> Is there a reason why the LSP setup computation
>>>> using
>>>>> RSVP-TE is done on the head end and not node by
>>>> node?
>>>>
>>>> As RSVP-TE uses (the information) from the IGP
>>>> topology it has the
>>>> information it needs to compute the path (and
>> match
>>>> it with the best
>>>> path request). Regardless, the path still has to
>> be
>>>> verified node by
>>>> node, and the nodes still have an opportunity to
>>>> decline the
>>>> reservation (which triggers an igp update). It
>>>> should be noted that
>>>> there are way to ways (loose ero) to distribute
>> the
>>>> path decision at
>>>> explicit places in the path.
>>>
>>>
>>>
>>> Say I wish to set up an LSP from A to B with the
>>> constraint that the number of hops should be x.
>>> Why is node by node computation not feasable in
>> this
>>> case?
>>
>> i'm not sure if the question is a theoretical one or
>> a practical one.
>> i'm guessing theoretical, afaik because a constraint
>> of 'number of
>> hops' isn't carried in the reservation
>>
>> unless you have something like ant routing, i don't
>> see how you can
>> define a constraint of hop-count and let the
>> decision be distributed
>> hop by hop. now if each node had a topology
>> database, it could make
>> that kind of decision hop-by-hop as long as the
>> constraint was carried
>> hop by hop as well. but if you already have a
>> topology database, why
>> try to do this computation hop-by-hop in the first
>> place? the
>> initiating node has all the information via it's
>> instance of the
>> topology database.
>
> Consider the case of an interarea or interas LSP
> setup,
> How exactly do people address this problem?
>
> I do not like to sit and write EROs for customer LSPs
> :((
> Is there a better method you can recommend?
>

there are several drafts that cover this area. i don't really know the 
current status though.


> -Sylvia
>
>
> 	
> 	
> 		
> ___________________________________________________________ALL-NEW 
> Yahoo! Messenger - all new features - even more fun!  
> http://uk.messenger.yahoo.com
>
>

-- 
Aamer Akhter / aa@cisco.com
NSITE cisco Systems

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml