--- 1/draft-ietf-mpls-ldp-interarea-02.txt 2008-02-26 00:12:37.000000000 +0100 +++ 2/draft-ietf-mpls-ldp-interarea-03.txt 2008-02-26 00:12:37.000000000 +0100 @@ -1,16 +1,16 @@ Network Working Group B. Decraene Internet Draft J.L. Le Roux - Document: draft-ietf-mpls-ldp-interarea-02.txt France Telecom - Expiration Date: August 2008 - I. Minei + Document: draft-ietf-mpls-ldp-interarea-03.txt France Telecom + Intended status: Standards Track + Expiration Date: August 2008 I. Minei Juniper Networks, Inc. February 2008 LDP extension for Inter-Area LSP Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware @@ -85,28 +85,27 @@ 3. Introduction Link state IGPs such as OSPF [OSPFv2] and IS-IS [IS-IS] allow the partition of an autonomous system into areas or levels so as to increase routing scalability within a routing domain. However, [LDP] requires that the IP address of the FEC Element should *exactly* match an entry in the IP RIB: according to [LDP] section 3.5.7.1 (Label Mapping Messages Procedures) "An LSR receiving a Label - Mapping message from a downstream LSR for a Prefix or Host Address - FEC Element should not use the label for forwarding unless its - routing table contains an entry that exactly matches the FEC - Element". + Mapping message from a downstream LSR for a Prefix SHOULD NOT use the + label for forwarding unless its routing table contains an entry that + exactly matches the FEC Element.". Therefore, MPLS LSPs between LERs in different areas/levels are not - setup unless the exact (/32 for IPv4) loopback addresses of all the - LERs are redistributed across all areas. + setup unless the specific (e.g. /32 for IPv4) loopback addresses of + all the LERs are redistributed across all areas. The problem statement is discussed in section 4. Then, in section 5 we extend the Label Mapping Procedure defined in [LDP] so as to support the setup of contiguous inter-area LSPs while maintaining IP prefix aggregation on the ABRs. This basically consists of allowing for "Longest Match Based" Label Mapping. 4. Problem statement Provider based MPLS VPN networks are expanding with the success of @@ -118,21 +117,21 @@ As a consequence many providers need to introduce IGP areas. Inter- area LSPs, that is LSPs that traverse at least two IGP areas, are required to ensure MPLS connectivity between PEs located in distinct IGP areas. To set up the required MPLS LSPs between PEs in different IGP areas, services providers have currently three solutions: LDP with IGP route leaking, BGP [MPLS-BGP] over LDP with MPLS hierarchy, or also inter- area RSVP-TE [ID-RSVP-TE]. - IGP route leaking consists in redistributing all /32 PE loopback + IGP route leaking consists in redistributing all specific PE loopback addresses across area boundaries. As a result, LDP finds in the RIB an exact match for its FEC and sets up the LSP. As a consequence, the potential benefits that a multi-area domain may yield are significantly diminished since a lot of addresses have to be redistributed by ABRs, and the number of IP entries in the LSDB and RIB maintained by every LSR of the domain (whatever the area/level it belongs to) cannot be minimized. Service providers may also set up these inter-area LSPs by using MPLS hierarchy with BGP [MPLS-BGP] as a label distribution protocol @@ -151,35 +150,37 @@ Service providers may also setup these inter-area LSPs by using inter-area RSVP-TE [ID-RSVP-TE]. This is a relevant solution when RSVP-TE is already used for setting up intra-area LSPs, and inter- area traffic engineering features are required. In return this is not a desired solution when LDP is already used for setting up intra-area LSPs, and inter-area traffic engineering features are not required. To avoid the above drawbacks, there is a need for an LDP based solution which allows setting up contiguous inter-area LSPs while - avoiding leaking of /32 PE loopback addresses across area boundaries, - and hence keeping all the benefits of IGP hierarchy. + avoiding leaking of specific PE loopback addresses across area + boundaries, and hence keeping all the benefits of IGP hierarchy. In that context, this document defines a new LDP Label Mapping Procedure so as to support the setup of contiguous inter-area LSPs while maintaining IP prefix aggregation on the ABRs. This procedure is similar to the one defined in [LDP] but performs a longest match when searching the FEC element in the RIB. 5. Label Mapping Procedure - This document defines a new label mapping procedure for LDP. It MUST - be possible to activate/deactivate this procedure by configuration - and it SHOULD be deactivated by default. It MAY be possible to - activate it on a per prefix basis. + This document defines a new label mapping procedure for [LDP]. It is + applicable to IPv4 and IPv6 prefix FEC elements (addresses families 1 + and 2 as per [ASSIGNED_AF]). It MUST be possible to activate / + deactivate this procedure by configuration and it SHOULD be + deactivated by default. It MAY be possible to activate it on a per + prefix basis. With this new longest match label mapping procedure, a LSR receiving a Label Mapping message from a neighbor LSR for a Prefix Address FEC Element (FEC1) SHOULD use the label for MPLS forwarding if its routing table contains an entry that matches the FEC Element and the advertising LSR is a next hop to reach the FEC. If so, it SHOULD advertise the received FEC Element (FEC1) and a label to its LDP peers. Note that this is the specific FEC (FEC1) that LDP re-advertise to its peers, and not the aggregated prefix found in the IP RIB during @@ -189,38 +190,37 @@ That is, either the LDP FEC element exactly matches an entry in the IP RIB or the FEC element is a subset of an IP RIB entry. There is no match for other cases such as the FEC element is a superset of a RIB entry. Note that with this longest match Label Mapping Procedure, each LSP established by LDP still strictly follows the shortest path(s) defined by the IGP. FECs selected by this "Longest Match" label mapping procedure are - distributed in an ordered way.. - In case of LER failure, the removal of reachability to the FEC occurs - using standard LDP procedures as defined in [LDP]. Similarly, the FEC - will be removed in an ordered way through the propagation of label - withdraw messages. The use of this reachability information by - application layers using the MPLS LSP (e.g., BGP) is outside the - scope of this document. + distributed in an ordered way. In case of LER failure, the removal of + reachability to the FEC occurs using standard LDP procedures as + defined in [LDP]. Similarly, the FEC will be removed in an ordered + way through the propagation of label withdraw messages. The use of + this (un)reachability information by application layers using the + MPLS LSP (e.g., BGP) is outside the scope of this document. As per [LDP], LDP has already some interactions with the RIB. In particular, it needs to be aware of the following events: - prefix UP when a new IP prefix appears in the RIB - prefix DOWN when an existing prefix disappears - next-hop change when an existing prefix have new next hop following a routing change. - With the longest match procedure, multiple FECs may be concerned by a - single RIB prefix change. The LSR must check all the FECs which are a - subset of this RIB prefix. So some LDP reactions following a RIB + With this longest match procedure, multiple FECs may be concerned by + a single RIB prefix change. The LSR must check all the FECs which are + a subset of this RIB prefix. So some LDP reactions following a RIB event are changed: - When a new prefix appears in the RIB, the LSR MUST check if this prefix is a better match for some existing FECs. E.g. the FEC elements 192.0.2.1/32 and 192.0.2.2/32 used the IP RIB entry 192.0.0/16 and a new more specific IP RIB entry 192.0.2/24 appears. This may result in changing the LSR used as next hop and hence the NHLFE for this FEC. - When a prefix disappears in the RIB, the LSR MUST check all FEC elements which are using this RIB prefix as best match. For each FEC, if another RIB prefix is found as best match, LDP MUST use @@ -255,32 +255,32 @@ +----------+ +-------------+ | | +--------------------------+ Figure 1: An IGP domain with two areas attached to the Backbone Area. Note that this applies equally to IS-IS and OSPF. An ABR refers here either to an OSPF ABR or to an IS-IS L1/L2 node. - All routers are MPLS enabled and MPLS connectivity (ie an LSP) is + All routers are MPLS enabled and MPLS connectivity (i.e. an LSP) is required between all PE routers. In the "egress" area "C", the records available are: IGP RIB LDP FEC elements: 192.0.2.1/32 192.0.2.1/32 192.0.2.2/32 192.0.2.2/32 192.0.2.3/32 192.0.2.3/32 The area border router ABR1 advertises in the backbone area: - the aggregated IP prefix 192.0.2/24 in the IGP - - all the individual IP FEC elements (/32) in LDP + - all the specific IP FEC elements (/32) in LDP In the "backbone" area "B", the records available are: IGP RIB LDP FEC elements: 192.0.2/24 192.0.2.1/32 192.0.2.2/32 192.0.2.3/32 The area border router ABR2 advertises in the area "A": - an aggregated IP prefix 192.0/16 in the IGP - all the individual IP FEC elements (/32) in LDP @@ -326,49 +326,51 @@ LSRs compliant with this document are backward compatible with LSRs that comply with [LDP]. For the successful establishment of end-to-end MPLS LSPs whose FEC are aggregated in the RIB, this specification must be implemented on all LSRs in all areas where IP aggregation is used. If an LSR on the path does not support this procedure, then the LSP initiated on the egress LSR stops at this non compliant LSR. There are no other adverse effects. - A service provider currently leaking specific (/32) LER's loopback + A service provider currently leaking specific LER's loopback addresses in the IGP and now considering performing IP aggregation on ABR should be aware that this may result in suboptimal routing as discussed in [RFC 2966]. - If all IP prefixes are leaked in the IGP backbone area and only stub - areas use IP aggregation, LSRs in the backbone area don't need to be - compliant with this document. + This extension does not necessarily require an AS wide deployment but + can be used on a per area basis. For example, if all specific IP + prefixes are leaked in the IGP backbone area and only stub areas use + IP aggregation, LSRs in the backbone area don't need to be compliant + with this document. 7.2. Impact on routing convergence time IGP convergence is based on two factors: the SPF calculation and FIB/LFIB update time. The SPF calculation scales O(N*Log(N)) where N is the number of Nodes. The FIB/LFIB update scales O(P) where P is the number of modified prefixes. Currently, with most routers implementations, the FIB/LFIB update is the dominant component [1]. - The solution documented in this draft reduces the Linkstate database + The solution documented in this draft reduces the link state database size and the number of FIB entries. However, it does not decrease the number of LFIB entries. In case of failure of the egress LER node, given that the IGP aggregates IP route on ABRs, the routing convergence behavior is changed compared to [LDP]. As the IGP does not carry specifics prefixes outside of the egress area, the IGP will not propagate the notification of node failure outside of the area. So application layers using the LSP may, as a local decision, use the availability of the MPLS LSP -as advertised by LDP- to achieve LER egress fault notification in addition to application layer fault detection - mechanisms. For some applications (e.g., BGP) this is expected to be + mechanisms. For some applications (e.g. BGP) this is expected to be faster. LDP will withdraw the FEC(s) of the egress LER in an ordered way through the end-to-end propagation of the LDP withdraw message. Compared to the IGP, this LDP failure notification may be faster or slower depending on the implementations, the IGP timers used and the network topology (network diameter). In typical topologies, the convergence time is expected to be similar or even faster since there is no need to wait for three consecutive SPF/PRC timers. For all others failures (links, Ps and ABRs nodes), the failure notification and the convergence are unchanged. @@ -383,20 +385,23 @@ This document has no actions for IANA. 10. References 10.1. Normative References [LDP] L. Andersson, I. Minei, B. Thomas, "LDP Specification", RFC 5036, October 2007 + [ASSIGNED_AF] http://www.iana.org/assignments/address-family- + numbers + 10.2. Informative References [L3-VPN] Rosen, E., Rekhter, Y. ," BGP/MPLS IP Virtual Private Networks (VPNs) ", RFC 4374, February 2006 [MPLS-BGP] Rekhter, Y., Rosen, E., "Carrying Label Information in [IS-IS] Callon, R., "Use of OSI IS-IS for Routing in TCP/IP and Dual Environments", RFC 1195, December 1990 [VPLS-BGP] Kompella, K., Rekhter, Y., "Virtual Private LAN Service @@ -418,21 +423,21 @@ second IGP convergence in large IP networks". ACM SIGCOMM 11. Acknowledgments Authors would like to thank Yakov Rekhter, Stefano Previdi, Vach Kompella, Bob Thomas, Clarence Filsfils, Kireeti Kompella, Luca Martini, Sina Mirtorabi, Dave McDysan, Benoit Fondeviole, Gilles Bourdon and Christian Jacquenet for the useful discussions on this subject, their review and comments. -Authors' Addresses +12. Authors' Addresses Bruno Decraene France Telecom 38 rue du General Leclerc 92794 Issy Moulineaux cedex 9 France EMail: bruno.decraene@orange-ftgroup.com Jean-Louis Le Roux @@ -441,21 +446,21 @@ 22307 Lannion Cedex France EMail: jeanlouis.leroux@orange-ftgroup.com Ina Minei Juniper Networks 1194 N. Mathilda Ave. Sunnyvale, CA 94089 - EMail: ina@juniper.net + Email: ina@juniper.net Intellectual Property Considerations The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be