Thread: [mpls-linux-general] Re: [MPLS-Linux] RSVP-TE (was: New mailing list)
Status: Beta
Brought to you by:
jleu
From: James R. L. <jl...@mi...> - 2000-11-30 17:59:59
|
On Thu, Nov 30, 2000 at 10:52:30PM +0530, Ramanjaneyulu Y T wrote: > Hi James , > This is really good job. I appreciate your efforts. Are u doing any work > related to RSVP-TE ?. Not as part of MPLS for Linux yet. I have done some work with RSVP-TE for my employeer. Because of that fact I can't be the one to start working on RSVP-TE for Linux, but if some else "starts" it I'll make it part of MPLS for Linux ;-) > I have one queastion related to LSP hierarchy. How LSRs distinguish > label stack ?. How remote LSR knows it label from other side , if > hierarchy is permited ?. First, I guess we need to get a few things sorted out, before I can answer that question. Are you refering to using RSVP-TE to create hierarchy, or are you refering to how hierarchy works in general. If your refering to using RSVP-TE to create hierarchy, are you confused about how to signal it or how (after it has been signaled) the ILM, NHLFE are setup to actually do hierarchy? Sorry to answer you question with questions. Jim BTW "First post" :-) -- James R. Leu |
From: Ramanjaneyulu Y T <yt...@cs...> - 2000-12-01 04:42:22
|
On Thu, 30 Nov 2000, James R. Leu wrote: > On Thu, Nov 30, 2000 at 10:52:30PM +0530, Ramanjaneyulu Y T wrote: > > Hi James , > > This is really good job. I appreciate your efforts. Are u doing any work > > related to RSVP-TE ?. > > Not as part of MPLS for Linux yet. I have done some work with RSVP-TE > for my employeer. Because of that fact I can't be the one to start working > on RSVP-TE for Linux, but if some else "starts" it I'll make it part > of MPLS for Linux ;-) > > > I have one queastion related to LSP hierarchy. How LSRs distinguish > > label stack ?. How remote LSR knows it label from other side , if > > hierarchy is permited ?. > > First, I guess we need to get a few things sorted out, before I can > answer that question. > > Are you refering to using RSVP-TE to create hierarchy, or are you refering > to how hierarchy works in general. I am referring to create hierarcht. > > If your refering to using RSVP-TE to create hierarchy, are you confused > about how to signal it or how (after it has been signaled) the ILM, NHLFE > are setup to actually do hierarchy? If I remember right , hierarchy is achieved with help of PUSH and FWD operations in ILM and NHLFE . Last night i refered some drafts and RSVP-TE supports Label stacks in LABEL object . My queastion is how LSR knows , how many labels pop ( i.e take off from) from LABEL object. Let me explain my probelm clealry. LER1 LER21 LER22 .... LER 2 |-----| |-----| . . | ----| |-----| | |----------| |------| |-------| | |-----| |-----| |-----| |-----| LER21 and LER22 is one MPLS domain. Suppose LER1 want to establish LSP to LER2. If one allows Hierarchy setting , it should make use of LSP between LER21 and LER22. when LER2 responded with RESV message, How this RESV message is treated in MPLS domain ( beteen LSR21 and LER22) ?. IN normal case each LSR gets one new label for previous hop and it binds this with next hop label , which he got from next hop. , then it propagets to previous hop. If we r setting hierarchial lSPs , only resource reseravtion done at each LSR when RESV messagae comes. No label storing abd getting new label at each core LSRs in MPLS domain 1 , except at edge routers. So how core LSrs will ignore to store label info ? . I think I am clear now. > > Sorry to answer you question with questions. > > Jim > > BTW "First post" :-) > |
From: James R. L. <jl...@mi...> - 2000-12-02 00:49:20
|
On Fri, Dec 01, 2000 at 10:11:49AM +0530, Ramanjaneyulu Y T wrote: <snip> > > Are you refering to using RSVP-TE to create hierarchy, or are you refering > > to how hierarchy works in general. > > I am referring to create hierarcht. > > > > > If your refering to using RSVP-TE to create hierarchy, are you confused > > about how to signal it or how (after it has been signaled) the ILM, NHLFE > > are setup to actually do hierarchy? > > > If I remember right , hierarchy is achieved with help of PUSH and FWD > operations in ILM and NHLFE . > > Last night i refered some drafts and RSVP-TE supports Label stacks in > LABEL object . My queastion is how LSR knows , how many labels pop ( i.e > take off from) from LABEL object. Let me explain my probelm clealry. When a LSR receives a labeled packet it extracts the label, bottotm of stack bit, ttl, and EXP bits from the first shim. It uses the label to lookup in the ILM table associated with this interface. The resulting ILM entry contains an instruction of what to do. If the operation is POP the LSR POPs the top shim. If the bottom of stack bit was set then the LSR looks to the L3 PID stored in the ILM entry. The ttl is coppied to the L3 header and the packet is sent to the L3 protocol for porcessing. If the bottom of stack bit was not set, then the label, the bottom os stack bit, ttl and EXP bits are exatrct from the next shim. The process repeats until either the bottom of stack bit is reached or an ILM entry with an intruction of FWD is found. > LER1 LER21 LER22 .... LER 2 > |-----| |-----| . . | ----| |-----| > | |----------| |------| |-------| | > |-----| |-----| |-----| |-----| > > > > > > LER21 and LER22 is one MPLS domain. Suppose LER1 want to > establish LSP to LER2. If one allows Hierarchy setting , it should make > use of LSP between LER21 and LER22. when LER2 responded with RESV message, > How this RESV message is treated in MPLS domain ( beteen LSR21 and LER22) > ?. IN normal case each LSR gets one new label for previous hop and it > binds this with next hop label , which he got from next hop. , then it > propagets to previous hop. > > If we r setting hierarchial lSPs , only resource reseravtion done at > each LSR when RESV messagae comes. No label storing abd getting new label > at each core LSRs in MPLS domain 1 , except at edge routers. So how core > LSrs will ignore to store label info ? . I'm a little unclear as to what you question is. I'll give it my best shot: |------------EDGE TO EGDE Tunneled LSP ---------| ------ ------ | | | | | PE | | PE | | 1 |---. ----- ----- .---| 2 | ------ \| | | |/ ------ | P |-- . . . . . . --| P | | 1 | | 2 | ----- ----- |----- CORE TE LSP -----| In the above picture the core of the network is traffic engineered with RSVP-TE with signalling starting at the P routers. The ER HOP lists are made up of the LSRIDs of each LSR the LSPs should traverse. The Edge to Egde LSPs could be setup with many protocols. So far I have only seen providers design networks that would use LDP to build the edge to egde LSPs. You asked about RSVP-TE I'm 100% sure, but I know the basic idea. RSVP-TE signalling is initiated at the PE routes. The ER HOP lists are composed of LSRIDs until it reaches the core of the network, and is supposed to tunnel inside of an existing LSP. At which point the ER HOP is the IP address associated with forwarding adjacency (term defined in draft-kompella-mpls-bundle). After thet the ER HOP list is again composed of LSRIDs. When RSVP-TE receives a PATH message that has a top ER HOP which is associated with a forwarding adjacency it is supposed to send the PATH message directly to the remote LSR associated with the forwarding adjacency via normal unicast routing. (I do not know all of the details of how that works or how the new LSR knows that this incoming PATH message is assocaited with an FA-LSP). When it comes time for the RESV message, the remote LSR allocates a label and sends it directly to this LSR via normal unicast routing. This LSR know that the RESV is associated with a forwarding adjacency (which is associated with an LSP) and installes the outgoing label with the instruction to PUSH the new label then PUSH on the label of the FA-LSP. I hope I didn't get it too wrong :-) Jim -- James R. Leu |