Re: [mpls-linux-general] Hierarchical LSP set up

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

I saw your post to the IETF mailing list.

Have you read the draft:

ftp://ftp.ietf.org/internet-drafts/draft-ietf-mpls-lsp-hierarchy-01.txt

It explains the notion of a Forwarding Adacency (FA).  When you establish
an LSP between two LSRs, you can consider them to have a FA, thus making the
LSP a FA-LSP.  The FA-LSP has an identifier associated with it (IP address
of some sort)  When signalling a LSP that should tunnel through the
FA-LSP, you signify this by contructing an ERO that has the IP address
assocaited with the FA-LSP as one hop.

RSVP-TE needs to then know how to send PATH and RESV messages to the far end
of a FA-LSP (I'm not sure how this works).

I hope this point you in the right direction.

Jim

PS When asking question to the IETF mailing list, the shorter
the question the better. :-)  Many of the people on the list
will not take the time to answer long questions.

On Wed, Dec 06, 2000 at 05:17:03PM +0530, Ramanjaneyulu Y T wrote:
> Hi,
> 
>    I have questions on how to establish LSPs . pls go through following
> mail and answer to those.
> 
>        Thanks in advance,
> 
> 
> 
> 
>   The following is the scenarios to establish hierarchial LSps.
> 
> Method1
> -------
> 
> 
> 
> 
> 
>     ---------						 ---------
>    |	     |						|	  |
>    | LER11   |						|  LER12  |
>    |	     |						|	  |
>     ---------					 	 ---------
> 	|						     |
> 	.       	  Y		      Z	   	     .
> 	.       --------       -------		  ------   T .
>         |__X___|        |     |       |          |      |____|
> 	       | LER1   |-----|       |-----.....|LER2  |
> 	       |	|     |       |          |      |
> 		--------       -------	          ------
>             
> at LER1                                       at LER2
> label look up table 				label look up table
>   		
>  X -- FWD Y					Z - FWD T
> 
> NHLFE    					NHLFE 	
> 
>  Y -- PUSH Y 					T -- PUSH T
> 
> 
> 
> 
> 
> Note:
> ----
>  PUSH y  -- PUSH label Y
>  FWD  y  -- check entry 'Y'
>             at NHLFE table    
> 
>   The output entry it having separate  operations. In this case it just
> forwads to soecified interface in table.
>                   
> 
>    ---
>     Initially there is no LSP existing between LER1 and LER2. When request
> comes from LER11 , PATH message is sent to LER12 . When LER12 responds
> with
> RESV message with label for previous hop.When RESV message reaches from
> upstream LSR to LER2 , it allocates new label and binds with label from
> upstream LSR and propagates RESV message to previous hop. The same process
> is
> repeated at all LSRS and LER1 also. Now LSP is established between from
> LER11
> to LER12 through LER1 and LER2. Note that this is not hierarchical LSP. 
> 
>   In this case , all data transfers occurs normally. 
> 
> Consider the following scenario.
> 
>   
> 
> 
>     --------						 ---------
>    |	     |						|	  |
>    | LER11   |						|  LER12  |
>    |	     |						|	  |
>     ---------					 	 ---------
> 	|						     |
> 	. 		  Y		  Z		     .
> 	.       --------       -------		  ------     .
>         |__X___|        |     |       |          |      |__T_|
> 	       | LER1   |-----|       |-----.....|LER2  |
> 	 ______|	|     |       |          |      |_____
> 	|A	--------       -------	          ------      | 	
> 	|						   B  |
> 	. 						      .	
> 	.						      .
> 							      |	
> 	|						 ----------
>      --------						|	   |
>     |	     |						|  LER 22  |
>     | LER 21 |						|	   |
>     |	     |						 ----------
>      -------- 					
> 
> 
> 
> In this scenario, we want to set up an LSP between LER21 and LER 22
> through
> LER1 and LER 2 which is a heirarchical LSP (in the sense that the same LSP
> wil
> be used by both the traffic flows from LER 11 and LER 21)
> 					
> Note:
> ----
>   LER2 generates label 'C' and informs to LER1 for LER21.  
> 
> 
> 
> at LER1                                       at LER2
> 
> label look up table 				label look up table
>   		
>  X -- FWD Y					Z - FWD T
>  A -- PUSH C , FWD Y                            C - FWD B
> 
> NHLFE    					NHLFE 	
> 
>  Y -- PUSH Y 					T -- PUSH T
> 						B -- PUSH B	
> 
> 
> 
> 
> 
> 
> 
>    If LER21 wants to establish LSP to LER22. It sends PATH request to
> LER22.
> LER22 responds to PATH message with RESV message with label for previous
> hop.
> when RESV message reaches to LER2 , it generates a new label for LER1 ,
> LSP
> is already existing between LER1 and LER2. In core LSRS between LER1 and
> LER2
> , just bypasses this message. At LER1 , it generates a new label and
> places
> PUSH LER2 label operation in newly generated label and forwards it to
> previous hop.
> 
> 
>     My question is , how LER2 differentiate MPLS data packets whether to
> forward
> to LER12 or LER22 by seeing on top label ( here 'Z'). 
> 
> Method2
> -------
>     If we establish one LSP between LER1 and LER2, then we can use this
> LSP to
> establish LER11 and LER12 as hierarchical LSPs. same way for LER21 and
> LER22.
> If this is the case at LER2 , data packet contains label stack of depth 2
> .
> So POP top label at LER2 and based on next label one can forward the
> packet. 
> But for this one is required to establish LSPs statically and those 2
> nodes
> should  know each other. But this approach is not suitable, i think , due
> to
> these constraints.
> 
> 
>        During LSP establishment between LER11 and LER12 , the LER2
> generates
> new label and informs to LER1. This is same for second LSP also.
> 
>     suppose LER2 generates M for LER12. The table update is as follows.
> 
> at LER1                                       at LER2
> 
> label look up table 				label look up table
>   		
>     						Z - POP
>  A -- PUSH C , FWD Y                            C - FWD B
>  X -- PUSH M , FWD Y  				M - FWD T
> 
> NHLFE    					NHLFE 	
> 
>  Y -- PUSH Y 					T -- PUSH T
> 						B -- PUSH B
> 								
> 
> 
>   So at LER2 , the MPLS pkt conatins M,Z or C,Z. So by looking on tables ,
> it
> learns POP top label and looks on next label. This is works fine in
> hierarchical environmant , but it has constraints.
> 
> Questions:
> ---------
> 
> 1. Is above methods are correct ?
> 
> 2.    Other than these 2 methods any other methods avaialble to achieve
> hierarchial LSPs ?.
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
>                                          
>                                          Regards
>                                         Ramanjaneyulu Y.T. 
> --  
>  " A real friend is one who walks in when the rest of the world walks
>  out."
>  ------------------------------------------------------------------------------ 
> Y.T.RAMANJANEYULU                        |   My other mail Ids:
> E-70,INDIAN INSTITUTE OF SCIENCE         |         yt...@12...
> BANGALORE - 560012                       |         ki...@ex...
> PH: 91 - 80 - 3092622 ( HOSTEL )         |
>     91 - 80 - 3092658 ( HFCL LAB )       |
>                    visit my home page:www2.csa.iisc.ernet.in/~ytr
> --------------------------------------------------------------------------------
> 
> 
> 
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> http://lists.sourceforge.net/mailman/listinfo/mpls-linux-general

-- 
James R. Leu