Network Working Group A. Bashandy, Ed.Internet Draft C. FilsfilsIntended status: Informational Cisco SystemsExpires: January 2025 P. Mohapatra Sproute Networks July 7, 2024BGP Prefix Independent Convergencedraft-ietf-rtgwg-bgp-pic-21.txtAbstractIn a network comprising thousands of BGP peers exchanging millions ofroutes, many routes are reachable via more than one next-hop. Giventhe large scaling targets, it is desirable to restore traffic afterfailure in a time period that does not depend on the number of BGPprefixes.This document describes an architecture by which traffic can be re-routed to equal cost multi-path (ECMP) or pre-calculated backup pathsin a timeframe that does not depend on the number of BGP prefixes.The objective is achieved through organizing the forwarding datastructures in a hierarchical manner and sharing forwarding elementsamong the maximum possible number of routes. The described techniqueyields prefix independent convergence while ensuring incrementaldeployment, complete automation, and zero management and provisioningeffort. It is noteworthy to mention that the benefits of BGP PrefixIndependent Convergence (BGP-PIC) are hinged on the existence of morethan one path whether as ECMP or primary-backup.Status of this Memo This Internet-Draft is submitted in full conformance with the provisions ofBCP 78 andBCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."Bashandy Expires January 7, 2025 [Page 1]
Internet-Draft BGP Prefix Independent Convergence July 2024 The list of current Internet-Drafts can be accessed athttp://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed athttp://www.ietf.org/shadow.html This Internet-Draft will expire on January 7, 2024.Copyright Notice Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject toBCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of theTrust Legal Provisions and are provided without warranty as described in the Simplified BSD License.Table of Contents1. Introduction...................................................31.1. Terminology...............................................32. Overview.......................................................62.1. Dependency................................................62.1.1. Hierarchical Hardware FIB (Forwarding Information Base) ............................................................62.1.2. Availability of more than one BGP next-hops..........72.2. BGP-PIC Illustration......................................73. Constructing the Shared Hierarchical Forwarding Chain.........103.1. Constructing the BGP-PIC Forwarding Chain................103.2. Example: Primary-Backup Pic-path Scenario................114. Forwarding Behavior...........................................125. Handling Platforms with Limited Levels of Hierarchy...........136. Forwarding Chain Adjustment at a Failure......................136.1. BGP-PIC core.............................................146.2. BGP-PIC edge.............................................156.2.1. Adjusting Forwarding Chain in egress node failure...156.2.2. Adjusting Forwarding Chain on PE-CE link Failure....156.3. Handling Failures for Flattened Forwarding Chains........177. Properties....................................................187.1. Coverage.................................................187.1.1. A remote failure on the pic-path to a BGP next-hop..187.1.2. A local failure on the pic-path to a BGP next-hop...18Bashandy Expires January 7, 2025 [Page 2]
Internet-Draft BGP Prefix Independent Convergence July 20247.1.3. A remote IBGP next-hop fails........................187.1.4. A local EBGP next-hop fails.........................187.2. Performance..............................................197.3. Automated................................................197.4. Incremental Deployment...................................198. Security Considerations.......................................209. IANA Considerations...........................................2010. References...................................................2010.1. Normative References....................................2010.2. Informative References..................................2011. Acknowledgments..............................................21Appendix A. Handling Platforms with Limited Levels of Hierarchy..23Appendix B. Example: Flattening a forwarding chain...............25Appendix C. Perspective..........................................321. Introduction BGP speakers exchange reachability information about prefixes [RFC4271] and, for labeled address families an edge router assigns local labels to prefixes and associates the local label with each advertised prefix using technologies such as L3VPN [RFC4364], 6PE [RFC4798], and Softwire [RFC5565] using BGP label unicast (BGP-LU) technique [RFC8277]. A BGP speaker then applies the path selection steps to choose the best route. In modern networks, it is not uncommon to have a prefix reachable via multiple edge routers. Multiple techniques have been described to allow for BGP to advertise more than one path for a given prefix [I.D.ietf-idr-best- external][RFC7911][RFC6774], whether in the form of equal cost multipath or primary-backup. Another common and widely deployed scenario is L3VPN with multi-homed VPN sites with unique Route Distinguisher. This document describes a hierarchical and shared forwarding chain organization that allows traffic to be restored to a pre- calculated alternative equal cost primary path or backup path in a time period that does not depend on the number of BGP prefixes. The technique relies on internal router behavior that is completely transparent to the operator and can be incrementally deployed and enabled with zero operator intervention. In other words, once it is implemented and deployed on a router, nothing is required from the operator to make it work. It is noteworthy to mention that this document describes a Forwarding Information Base (FIB) architecture that can be implemented in both hardware and/or software, although we refer to hardware implementation in most of the cases because of the additional complexity and performance requirements associated with hardware implementations.1.1. Terminology This section defines the terms used in this document.Bashandy Expires January 7, 2025 [Page 3]
Internet-Draft BGP Prefix Independent Convergence July 2024 o BGP-LU: BGP Label Unicast. Refers to carrying label unicast address family (SAFI-4) in BGP4 as in [RFC8277]. o BGP prefix: A IP address prefix as described in [RFC4271]. o IGP prefix: A prefix that is learnt via an Interior Gateway Protocol (IGP), such as OSPF and ISIS. The prefix may be learnt directly through the IGP or statically configured. o Customer Edge (CE) [RFC4364]: An external router through which an egress PE can reach a prefix P/m. o Egress PE [RFC4364], "ePE": A BGP speaker that learns about a prefix through an external BGP (EBGP) peer and chooses that EBGP peer as the next-hop for that prefix. o Ingress PE, "iPE": A BGP speaker that learns about a prefix through a Internal BGP (IBGP) peer and chooses an egress PE as the next-hop for the prefix. o Pic-path: The next-hop in a sequence of nodes starting from the current node and ending with the destination node or network identified by the prefix. The nodes may not be directly connected. o Recursive pic-path: A pic-path consisting only of the IP address of the next-hop without the outgoing interface. Subsequent lookups are necessary to determine the outgoing interface and a directly connected next-hop. o Non-recursive pic-path: A pic-path consisting of the IP address of a directly connected next-hop and outgoing interface. o Adjacency: The layer 2 encapsulation leading to the layer 3 directly connected next-hop. An adjacency is identified by a next-hop and an outgoing interface o Primary pic-path: A recursive or non-recursive pic-path that can be used for forwarding as long as forwarding engine can walk (Seesection 2.2 for explanation of forwarding chain andSection 4 forwarding engine behavior) starting from this pic- path can end to an adjacency. A prefix can have more than one primary pic-path. o Backup pic-path: A recursive or non-recursive pic-path that can be used only after some or all primary pic-paths become unreachable. o Primary Next-hop. The next-hop in a primary pic-pathBashandy Expires January 7, 2025 [Page 4]
Internet-Draft BGP Prefix Independent Convergence July 2024 o Secondary next-hop: The next-hop in the backup pic-path o Leaf: A container data structure for a prefix or local label. Alternatively, it is the data structure that contains prefix specific information. o IP leaf: The leaf corresponding to an IPv4 or IPv6 prefix. o Label leaf. The leaf corresponding to a locally allocated label such as the VPN label on an egress PE [RFC4364]. o Pathlist: An array of pic-paths used by one or more prefixes to forward traffic to destination(s) covered by an IP prefix. Each pic-path in the pathlist carries its "path-index" that identifies its position in the array of paths. In general, the value of the path-index in a pic-path is the same as its position in the pathlist, except in the case outlined inSection 5. For example the 3rd pic-path may carry a path-index value of 1. A pathlist may contain a mix of primary and backup pic-paths. o OutLabel-List: Each labeled prefix is associated with an OutLabel-List. The OutLabel-List is an array of one or more outgoing labels and/or label actions where each label or label action has 1-to-1 correspondence to a pic-path in the pathlist. Label actions are: push (add) the label as specified in [RFC3031], pop (remove) the label as specified in [RFC3031], swap (replace) the incoming label with the label in the OutLabel-List entry, or don't push anything at all in case of "unlabeled". The prefix may be an IGP or BGP prefix. o Forwarding chain: It is a compound data structure consisting of multiple connected blocks that a forwarding engine walks one block at a time to forward the packet out of an interface.Section 2.2 explains an example of a forwarding chain. Subsequent sections provide additional examples o Dependency: An object X is said to be a dependent or child of object Y if there is at least one forwarding chain where the forwarding engine must visit the object X before visiting the object Y in order to forward a packet. Note that if object X is a child of object Y, then Y cannot be deleted unless object X is no longer a dependent/child of object Y. o Pic-route: A prefix with one or more pic-paths associated with it. The minimum set of objects needed to construct a pic-route is a leaf and a pathlist. o IGP pic-route: a pic-route whose prefix is learned from an IGP o BGP pic-route: a pic-route whose prefix is learned from BGPBashandy Expires January 7, 2025 [Page 5]
Internet-Draft BGP Prefix Independent Convergence July 2024 o Routing-table: A table where each entry is a pic-route as defined in this section. o ASN: Autonomous System Number2. Overview The idea of BGP-PIC is based on two pillars o A shared hierarchical forwarding chain: It is not uncommon to see multiple destinations reachable via the same list of next-hops. Instead of having a separate list of next-hops for each destination, all destinations sharing the same list of next-hops can point to a single copy of this list thereby allowing fast convergence by making changes to a single shared list of next- hops rather than possibly a large number of destinations. Because pic-paths in a pathlist may be recursive, a hierarchy is formed between pathlist and the resolving prefix whereby the pathlist depends on the resolving prefix. o A forwarding plane that supports multiple levels of indirection: A forwarding chain that starts with a destination and ends with an outgoing interface is not a simple flat structure. Instead, a forwarding entry is constructed via multiple levels of indirections. A BGP prefix uses a recursive next-hop, which in turn resolves via an IGP next-hop, which in turn resolves via an adjacency consisting of one or more outgoing interface(s) and next-hop(s). Designing a forwarding plane that constructs multi-level forwarding chains with maximal sharing of forwarding objects allows rerouting a large number of destinations by modifying a small number of objects thereby achieving convergence in a time frame that does not depend on the number of destinations. For example, if the IGP prefix that resolves a recursive next-hop is updated there is no need to update the possibly large number of BGP NLRIs that use this recursive next- hop.2.1. Dependency This section describes the required functionalities in the forwarding and control planes to support BGP-PIC as described in this document.2.1.1. Hierarchical Hardware FIB (Forwarding Information Base) BGP-PIC requires a hierarchical hardware FIB support: if the destination address of a forwarded packet matches a BGP prefix, aBashandy Expires January 7, 2025 [Page 6]
Internet-Draft BGP Prefix Independent Convergence July 2024 BGP leaf is looked up, then a BGP pathlist is consulted, then an IGP pathlist, then an adjacency.Section 4 has more details about how a packet is forwarded An alternative method consists in "flattening" the dependencies when programming the BGP destinations into HW FIB resulting in potentially eliminating both the BGP pathlist and IGP pathlist consultation. Such an approach decreases the number of memory lookups per forwarding operation at the expense of HW FIB memory increase (flattening means less sharing thereby less duplication), loss of equal cost multi-path (ECMP) properties (flattening means less pathlist entropy) and loss of BGP-PIC properties.Section 5 explains the concept of flattening for hardware with limited number of levels of indirections.2.1.2. Availability of more than one BGP next-hops When the BGP next-hop in the primary pic-path becomes unresolved, BGP-PIC depends on the availability of one or more pre-computed and pre-programmed backup pic-paths(s) in the BGP pathlist in the forwarding engine. The existence of a backup pic-path is clearly required for the following reason: a network connectivity service caring for network availability will require two disjoint network connections resulting in two BGP next-hops. The BGP distribution of secondary next-hops is available thanks to the following BGP mechanisms: Add-Path [RFC7911], BGP Best-External [I.D.ietf-idr-best-external], diverse path [RFC6774], and the frequent use in VPN deployments of different VPN RD's per PE. Another option to learn multiple BGP next-hops/paths is to receive IBGP paths from multiple BGP RRs [RFC9107] selecting a different path as best. It is noteworthy to mention that the availability of another BGP path does not mean that all failure scenarios can be covered by simply forwarding traffic to the available secondary path. The discussion of how to cover various failure scenarios is beyond the scope of this document.2.2. BGP-PIC Illustration To illustrate the two pillars above as well as the platform dependency, this document will use an example of a multihomed L3VPN prefix in a BGP-free core running LDP [RFC5036] or segment routing over MPLS forwarding plane [RFC8660].Bashandy Expires January 7, 2025 [Page 7]
Internet-Draft BGP Prefix Independent Convergence July 2024 +--------------------------------+ | | | ePE2 (IGP-IP1 192.0.2.1, Loopback) | | \ | | \ | | \ iPE | CE....VRF "Blue", ASN 65000 | | / (VPN-IP1 198.51.100.0/24) | | / (VPN-IP2 203.0.113.0/24) | LDP/Segment-Routing Core | / | ePE1 (IGP-IP2 192.0.2.2, Loopback) | | +--------------------------------+ Figure 1: VPN prefix reachable via multiple PEs Referring to Figure 1, suppose the iPE (the ingress PE) receives NLRIs for the VPN prefixes VPN-IP1 and VPN-IP2 from two egress PEs, ePE1 and ePE2 with next-hop BGP-NH1 (192.0.2.1) and BGP-NH2 (192.0.2.2), respectively. Assume that ePE1 advertise the VPN labels VPN-L11 and VPN-L12 while ePE2 advertise the VPN labels VPN-L21 and VPN-L22 for VPN-IP1 and VPN-IP2, respectively. Suppose that BGP-NH1 and BGP-NH2 are resolved via the IGP prefixes IGP-IP1 and IGP-IP2, where each happen to have 2 equal cost paths with IGP-NH1 and IGP- NH2 reachable via the interfaces I1 and I2 on iPE, respectively. Suppose that local labels (whether LDP [RFC5036] or segment routing [RFC8660]) on the downstream LSRs for IGP-IP1 are IGP-L11 and IGP- L12 while for IGP-IP2 are IGP-L21 and IGP-L22. As such, the pic- routing table at iPE is as follows: 65000: 198.51.100.0/24 via ePE1 (192.0.2.1), VPN Label: VPN-L11 via ePE2 (192.0.2.2), VPN Label: VPN-L21 65000: 203.0.113.0/24 via ePE1 (192.0.2.1), VPN Label: VPN-L12 via ePE2 (192.0.2.2), VPN Label: VPN-L22 192.0.2.1/32 (ePE2) via I1, Label: IGP-L11 via I2, Label: IGP-L12 192.0.2.2/32 (ePE1) via I1, Label: IGP-L21 via I2, Label: IGP-L22 Based on the above pic-routing-table, a hierarchical forwarding chain can be constructed as shown in Figure 2.Bashandy Expires January 7, 2025 [Page 8]
Internet-Draft BGP Prefix Independent Convergence July 2024 IP Leaf: pathlist: IP Leaf: pathlist: -------- +-----------+ -------- | | +-------------+ |BGP-NH1------->IGP-IP1 ----->| | VPN-IP1-->| | | | IGP-NH1,I1----->adjacency1 | |BGP-NH2------->... | | | | | | | | IGP-NH2,I2----->adjacency2 | +-----------+ | | | | | +-------------+ | | v v OutLabel-List: OutLabel-List: +--------+ +--------+ |VPN-L11 | |IGP-L11 | |VPN-L21 | |IGP-L12 | +--------+ +--------+ Figure 2: Shared Hierarchical Forwarding Chain at iPE The forwarding chain depicted in Figure 2 illustrates the first pillar, which is sharing and hierarchy. It can be seen that the BGP pathlist consisting of BGP-NH1 and BGP-NH2 is shared by all NLRIs reachable via ePE1 and ePE2. As such, it is possible to make changes to the pathlist without having to make changes to the NLRIs. For example, if BGP-NH2 becomes unreachable, there is no need to modify any of the possibly large number of NLRIs. Instead only the shared pathlist needs to be modified. Likewise, due to the hierarchical structure of the forwarding chain, it is possible to make modifications to the IGP pic-routes without having to make any changes to the BGP NLRIs. For example, if the interface "I2" goes down, only the shared IGP pathlist needs to be updated, but none of the IGP prefixes sharing the IGP pathlist nor the BGP NLRIs using the IGP prefixes for resolution need to be modified. Figure 2 can also be used to illustrate the second BGP-PIC pillar. Having a deep forwarding chain such as the one illustrated in Figure 2 requires a forwarding plane that is capable of accessing multiple levels of indirection in order to calculate the outgoing interface(s) and next-hops(s). While a deeper forwarding chain minimizes the re-convergence time on topology change, there will always exist platforms with limited capabilities and hence imposing a limit on the depth of the forwarding chain.Section 5 describes how to gracefully trade off convergence speed with the number of hierarchical levels to support platforms with different capabilities. Another example using IPv6 addresses can be something like the followingBashandy Expires January 7, 2025 [Page 9]
Internet-Draft BGP Prefix Independent Convergence July 2024 65000: 2001:DB8:1::/48 via ePE1 (65000: 2001:DB8:192::1), VPN Label: VPN6-L11 via ePE2 (65000: 2001:DB8:192::2), VPN Label: VPN6-L21 65000: 2001:DB8:2:/48 via ePE1 (65000: 2001:DB8:192::1), VPN Label: VPN6-L12 via ePE2 (65000: 2001:DB8:192::2), VPN Label: VPN6-L22 65000: 2001:DB8:192::1/128 via Core, Label: IGP6-L11 via Core, Label: IGP6-L12 65000: 2001:DB8:192::2/128 via Core, Label: IGP6-L21 via Core, Label: IGP6-L22 The same hierarchical forwarding chain described can be constructed for IPv6 addresses/prefixes.3. Constructing the Shared Hierarchical Forwarding Chain Constructing the forwarding chain is an application of the two pillars described inSection 2. This section describes how to construct the forwarding chain in a hierarchical shared manner.3.1. Constructing the BGP-PIC Forwarding Chain The whole process starts when a BGP prefix is downloaded to FIB. The prefix contains one or more outgoing pic-paths. For certain labeled prefixes, such as L3VPN [RFC4364] prefixes, each pic-path may be associated with an outgoing label and the prefix itself may be assigned a local label. The list of outgoing pic-paths defines a pathlist. If such pathlist does not already, then the FIB manager (software or hardware entity responsible for managing the FIB) creates a new pathlist, otherwise the existing pathlist with the same list of pic-paths exist (the pathlist may already exist because there is another pic-route that is already using the same list of pic-paths) is used. The BGP prefix is added as a dependent of the pathlist. The previous step constructs the upper part of the hierarchical forwarding chain. The forwarding chain is completed by resolving the pic-paths of the pathlist. A BGP pic-path usually consists of a next-hop. The next-hop is resolved by finding a matching IGP prefix. The end result is a hierarchical shared forwarding chain where the BGP pathlist is shared by all BGP prefixes that use the same list ofBashandy Expires January 7, 2025 [Page 10]
Internet-Draft BGP Prefix Independent Convergence July 2024 pic-paths and the IGP prefix is shared by all pathlists that have a pic-path resolving via that IGP prefix. The remainder of this section goes over an example to illustrate the applicability of BGP-PIC in a primary-backup pic-path scenario.3.2. Example: Primary-Backup Pic-path Scenario Consider the egress PE ePE1 in the case of the multi-homed VPN prefixes shown in Figure 1. Suppose ePE1 determines that the primary pic-path is the external pic-path, while the backup pic-path is the IBGP pic-path to the other PE ePE2 with next-hop BGP-NH2. ePE1 constructs the forwarding chain depicted in Figure 3. The figure shows only a single VPN prefix for simplicity. But all prefixes that are multihomed to ePE1 and ePE2 share the BGP pathlist. BGP OutLabel-List +---------+ VPN-L11 |Unlabeled| (Label-leaf)---+---->+---------+ | | VPN-L21 | v | (swap) | | +---------+ | | | | | BGP pathlist | +--------------+ | | | | | CE-NH ------->(to the CE) | | path-index=0 | VPN-IP1 -----+------------------>+--------------+ (IP leaf) | VPN-NH2 | | | (backup) ------->IGP Leaf | | path-index=1 | (Towards ePE2) | +--------------+ | | BGP OutLabel-List | +---------+ | |Unlabeled| +------------->+---------+ | VPN-L21 | | (push) | +---------+ Figure 3: VPN Prefix Forwarding Chain with eiBGP pic-paths on egress PEBashandy Expires January 7, 2025 [Page 11]
Internet-Draft BGP Prefix Independent Convergence July 2024 The example depicted in Figure 3 differs from the example in Figure 2 in two main aspects. First, as long as the primary pic-path towards the CE (external pic-path) can be used for forwarding, it will be the only pic-path used for forwarding while the OutLabel- List contains both the unlabeled (primary pic-path) and the VPN label (backup pic-path) advertised by the backup pic-path ePE2. The second aspect is presence of the label leaf corresponding to the VPN prefix. This label leaf is used to match VPN traffic arriving from the core. Note that the label leaf shares the pathlist with the IP prefix.4. Forwarding Behavior This section explains how the forwarding plane uses the hierarchical shared forwarding chain to forward a packet. When a packet arrives at a router, assume it matches a leaf. If not, the packet is handled according to the local policy (such as silently dropping the packet), which is beyond the scope of this document. A labeled packet matches a label leaf while an IP packet matches an IP leaf. The forwarding engines walks the forwarding chain starting from the leaf until the walk terminates on an adjacency. Thus when a packet arrives, the chain is walked as follows: 1. Lookup the leaf based on the destination address or the label at the top of the packet. 2. Retrieve the parent pathlist of the leaf. 3. Pick an outgoing pic-path "Pi" from the list of resolved pic- paths in the pathlist. The method by which the outgoing pic-path is picked is beyond the scope of this document (e.g. flow- preserving hash exploiting entropy within the MPLS stack and IP header). Let the "path-index" of the outgoing pic-path "Pi" be "j". Remember that, as described in the definition of the term pathlist inSection 1.1, the path-index of a pic-path may not always be identical the position of the pic-path in the pathlist. 4. If the prefix is labeled, use the "path-index" "j" to retrieve the label "Lj" stored position j in the OutLabel-List and apply the label action of the label on the packet (e.g. for VPN label on the ingress PE, the label action is "push"). As mentioned inSection 1.1 the value of the "path-index" stored in the pic- path may not necessarily be the same value of the location of the pic-path in the pathlist.Bashandy Expires January 7, 2025 [Page 12]
Internet-Draft BGP Prefix Independent Convergence July 2024 5. If the chosen pic-path "Pi" is recursive, move to its parent prefix and go to step 2. 6. If the chosen pic-path is non-recursive move to its parent adjacency. 7. Encapsulate the packet in the layer string specified by the adjacency and send the packet out. Let's apply the above forwarding steps to the forwarding chain depicted in Figure 2 inSection 2. Suppose a packet arrives at ingress PE iPE from an external neighbor. Assume the packet matches the VPN prefix VPN-IP1. While walking the forwarding chain, the forwarding engine applies a hashing algorithm to choose the pic-path and the hashing at the BGP level chooses the first pic-path in the BGP pathlist while the hashing at the IGP level yields the second pic-path in the IGP pathlist. In that case, the packet will be sent out of interface I2 with the label stack "IGP-L12,VPN-L11".5. Handling Platforms with Limited Levels of Hierarchy This section describes the construction of the forwarding chain if a platform does not support the number of recursion levels required to resolve the NLRIs. There are two main design objectives. o Being able to reduce the number of hierarchical levels from any arbitrary value to a smaller arbitrary value that can be supported by the forwarding engine. o Minimal modifications to the forwarding algorithm due to such reduction.Appendix A provides details on how to handle limited hardware capabilities.6. Forwarding Chain Adjustment at a Failure The hierarchical and shared structure of the forwarding chain explained in the previous section allows modifying a small number of forwarding chain objects to re-route traffic to a pre-calculated equal-cost or backup pic-path without the need to modify the possibly very large number of BGP prefixes. This section goes over various core and edge failure scenarios to illustrate how the FIB manager can utilize the forwarding chain structure to achieve BGP prefix independent convergence.Bashandy Expires January 7, 2025 [Page 13]
Internet-Draft BGP Prefix Independent Convergence July 20246.1. BGP-PIC core This section describes the adjustments to the forwarding chain when a core link or node fails but the BGP next-hop remains reachable. There are two case: remote link failure and attached link failure. Node failures are treated as link failures. When a remote link or node fails, the IGP on the ingress PE receives an advertisement indicating a topology change so IGP re-converges to either find a new next-hop and/or outgoing interface or remove the pic-path completely from the IGP prefix used to resolve BGP next- hops. IGP and/or LDP download the modified IGP leaves with modified outgoing labels for the labeled core. When a local link fails, FIB manager detects the failure almost immediately. The FIB manager marks the impacted pic-path(s) as unusable so that only useable pic-paths are used to forward packets. Hence only IGP pathlists with pic-paths using the failed local link need to be modified. All other pathlists are not impacted. Note that in this particular case there is no need to backwalk (walk back the forwarding chain) to IGP leaves to adjust the OutLabel-Lists because FIB can rely on the path-index stored in the useable pic-paths in the pathlist to pick the right label. It is noteworthy to mention that because FIB manager modifies the forwarding chain starting from the IGP leaves only. BGP pathlists and leaves are not modified. Hence traffic restoration occurs within the time frame of IGP convergence, and, for local link failure, assuming a backup pic-path has been precomputed, within the timeframe of local detection (e.g. 50ms). Examples of solutions that can pre-compute backup pic-paths are IP FRR [RFC5714] remote LFA [RFC7490], TI-LFA [I-D.ietf-rtgwg-segment-routing-ti-lfa] and MRT [RFC7812] or EBGP pic-path having a backup pic-path [bonaventure]. Let's apply the procedure mentioned in this subsection to the forwarding chain depicted in Figure 2. Suppose a remote link failure occurs and impacts the first ECMP IGP pic-path to the remote BGP next-hop. Upon IGP convergence, the IGP pathlist used by the BGP next-hop is updated to reflect the new topology (one pic-path instead of two) and the new forwarding state is immediately available to all dependent BGP prefixes. The same behavior would occur if the failure was local such as an interface going down. As soon as the IGP convergence is complete for the BGP next-hop IGP pic-route, all its BGP depending routes benefit from the new pic- path. In fact, upon local failure, if LFA protection is enabled for the IGP pic-route to the BGP next-hop and a backup pic-path was pre- computed and installed in the pathlist, upon the local interface failure, the LFA backup pic-path is immediately activated (e.g. sub-Bashandy Expires January 7, 2025 [Page 14]
Internet-Draft BGP Prefix Independent Convergence July 2024 50msec) and thus protection benefits all the depending BGP traffic through the hierarchical forwarding dependency between the routes.6.2. BGP-PIC edge This section describes the adjustments to the forwarding chains as a result of edge node or edge link failure.6.2.1. Adjusting Forwarding Chain in egress node failure When a node fails, IGP on neighboring core nodes send updates indicating that the edge node is no longer a direct neighbor. If the node that failed is an egress node, such as ePE1 and ePE2 in Figure 1, IGP running on an ingress node, such as iPE in Figure 1, converges and the realizes that the egress node is no longer reachable. As such IGP on the ingress node instructs FIB to remove the IP and label leaves corresponding to the failed edge node from FIB. So FIB manager on the ingress node performs the following steps: o FIB manager deletes the IGP leaf corresponding to the failed edge node o FIB manager backwalks to all dependent BGP pathlists and marks that pic-path using the deleted IGP leaf as unresolved o Note that there is no need to modify the possibly large number of BGP leaves because each pic-path in the pathlist carries its pic- path index and hence the correct outgoing label will be picked. Consider for example the forwarding chain depicted in Figure 2. If the 1st BGP pic-path becomes unresolved, then the forwarding engine will only use the second pic-path for forwarding. Yet the path-index of that single resolved pic-path will still be 1 and hence the label VPN-L21 will be pushed.6.2.2. Adjusting Forwarding Chain on PE-CE link Failure Suppose the link between an edge router and its external peer fails. There are two scenarios (1) the edge node attached to the failed link performs next-hop self (where BGP advertises the IP address of its own loopback as next-hop) and (2) the edge node attached to the failure advertises the IP address of the failed link as the next-hop attribute to its IBGP peers. In the first case, the rest of IBGP peers will remain unaware of the link failure and will continue to forward traffic to the edge node until the edge node attached to the failed link withdraws the BGP prefixes. If the destination prefixes are multi-homed to anotherBashandy Expires January 7, 2025 [Page 15]
Internet-Draft BGP Prefix Independent Convergence July 2024 IBGP peer, say ePE2, then FIB manager on the edge router detecting the link failure applies the following steps to the forwarding chain (see Figure 3): o FIB manager backwalks to the BGP pathlists marks the pic-path through the failed link to the external peer as unresolved. o Hence traffic will be forwarded using the backup pic-path towards ePE2. o Labeled traffic arriving at the egress PE ePE1 matches the BGP label leaf. o The OutLabel-List attached to the BGP label leaf already contains an entry corresponding to the backup pic-path. o The label entry in OutLabel-List corresponding to the internal pic-path to backup egress PE has a swap action to the label advertised by the backup egress PE. o For an arriving label packet (e.g. VPN), the top label is swapped with the label advertised by backup egress PE and the packet is sent towards that the backup egress PE. o Unlabeled traffic arriving at the egress PE ePE1 matches the BGP IP leaf o The OutLabel-List attached to the BGP label leaf already contains an entry corresponding to the backup pic-path. o The label entry in OutLabel-List corresponding to the internal pic-path to backup egress PE has a push (instead of the swap action in for the labeled traffic case) action to the label advertised by the backup egress PE. o For an arriving IP packet, the label advertised by backup egress PE is pushed and the packet is sent towards that the backup egress PE. In the second case where the edge router uses the IP address of the failed link as the BGP next-hop, the edge router will still perform the previous steps. But, unlike the case of next-hop self, the IGP on the failed edge node informs the rest of the IBGP peers that the IP address of the failed link is no longer reachable. Hence the FIB manager on IBGP peers will delete the IGP leaf corresponding to the IP prefix of the failed link. The behavior of the IBGP peers will be identical to the case of edge node failure outlined inSection6.2.1.Bashandy Expires January 7, 2025 [Page 16]
Internet-Draft BGP Prefix Independent Convergence July 2024 It is noteworthy to mention that because the edge link failure is local to the edge router, sub-50 msec convergence can be achieved as described in [bonaventure]. Let's try to apply the case of next-hop self to the forwarding chain depicted in Figure 3. After failure of the link between ePE1 and CE, the forwarding engine will route traffic arriving from the core towards VPN-NH2 with path-index=1. A packet arriving from the core will contain the label VPN-L11 at top. The label VPN-L11 is swapped with the label VPN-L21 and the packet is forwarded towards ePE2.6.3. Handling Failures for Flattened Forwarding Chains As explained in the inSection 5 if the number of hierarchy levels of a platform cannot support the native number of hierarchy levels of a recursive forwarding chain, the instantiated forwarding chain is constructed by flattening two or more levels. Hence a 3-levels chain in Figure 5 is flattened into the 2-levels chain in Figure 6. While reducing the benefits of BGP-PIC, flattening one hierarchy into a shallower hierarchy does not always result in a complete loss of the benefits of the BGP-PIC. To illustrate this fact suppose ASBR12 is no longer reachable in domain 1. If the platform supports the full hierarchy depth, the forwarding chain is the one depicted in Figure 5 and hence the FIB manager needs to backwalk one level to the pathlist shared by "ePE1" and "ePE2" and adjust it. If the platform supports 2 levels of hierarchy, then a useable forwarding chain is the one depicted in Figure 6. In that case, if ASBR12 is no longer reachable, the FIB manager has to backwalk to the two flattened pathlists and updates both of them. The main observation is that the loss of convergence speed due to the loss of hierarchy depth depends on the structure of the forwarding chain itself. To illustrate this fact, let's take two extremes. Suppose the forwarding objects in level i+1 depend on the forwarding objects in level i. If every object on level i+1 depends on a separate object in level i, then flattening level i into level i+1 will not result in loss of convergence speed. Now let's take the other extreme. Suppose "n" objects in level i+1 depend on 1 object in level i. Now suppose FIB flattens level i into level i+1. If a topology change results in modifying the single object in level i, then FIB has to backwalk and modify "n" objects in the flattened level, thereby losing all the benefit of BGP-PIC. Experience shows that flattening forwarding chains usually results in moderate loss of BGP-PIC benefits. Further analysis is needed to corroborate and quantify this statement.Bashandy Expires January 7, 2025 [Page 17]
Internet-Draft BGP Prefix Independent Convergence July 20247. Properties7.1. Coverage All the possible failures, except CE node failure, are covered, whether they impact a local or remote IGP pic-path or a local or remote BGP next-hop as described inSection 6. This section provides details for each failure and how the hierarchical and shared FIB structure described in this document allows recovery that does not depend on number of BGP prefixes.7.1.1. A remote failure on the pic-path to a BGP next-hop Upon IGP convergence, the IGP leaf for the BGP next-hop is updated and all the BGP depending routes leverage the new IGP forwarding state immediately. Details of this behavior can be found inSection6.1. This results in BGP traffic recovery that only depends on IGP convergence and is independent of the number of BGP prefixes impacted.7.1.2. A local failure on the pic-path to a BGP next-hop Upon LFA protection, the IGP leaf for the BGP next-hop is updated to use the precomputed backup pic-path and all the BGP depending routes leverage this protection. Details of this behavior can be found inSection 6.1. This BGP resiliency property only depends on LFA protection and is independent of the number of BGP prefixes impacted.7.1.3. A remote IBGP next-hop fails Upon IGP convergence, the IGP leaf for the BGP next-hop is deleted and all the depending BGP Path-Lists are updated to either use the remaining ECMP BGP best-paths or if none remains available to activate precomputed backups. Details about this behavior can be found inSection 6.2.1. This BGP resiliency property only depends on IGP convergence and is independent of the number of BGP prefixes impacted.7.1.4. A local EBGP next-hop fails Upon local link failure detection, the adjacency to the BGP next-hop is deleted and all the depending BGP pathlists are updated to either use the remaining ECMP BGP best-paths or if none remains availableBashandy Expires January 7, 2025 [Page 18]
Internet-Draft BGP Prefix Independent Convergence July 2024 to activate precomputed backups. Details about this behavior can be found inSection 6.2.2. This BGP resiliency property only depends on local link failure detection and is independent of the number of BGP prefixes impacted.7.2. Performance When the failure is local (a local IGP next-hop failure or a local EBGP next-hop failure), a pre-computed and pre-installed backup is activated by a local-protection mechanism that does not depend on the number of BGP destinations impacted by the failure. Sub-50msec is thus possible even if millions of BGP prefixes are impacted. When the failure is remote (a remote IGP failure not impacting the BGP next-hop or a remote BGP next-hop failure), an alternate pic- path is activated upon IGP convergence. All the impacted BGP destinations benefit from a working alternate pic-path as soon as the IGP convergence occurs for their impacted BGP next-hop even if millions of BGP pic-routes are impacted.Appendix A puts the BGP-PIC benefits in perspective by providing some results using actual numbers.7.3. Automated The BGP-PIC solution does not require any operator involvement. The process is entirely automated as part of the FIB implementation. The salient points enabling this automation are: o Extension of the BGP Best path to compute more than one primary ([RFC7911] and [RFC6774]) or backup BGP next-hop ([I.D.ietf-idr- best-external] and [I-D.pmohapat-idr-fast-conn-restore]). o Sharing of BGP Pathlist across BGP destinations with the same primary and backup BGP next-hop. o Hierarchical indirection and dependency between BGP pathlist and IGP pathlist.7.4. Incremental Deployment As soon as one router supports BGP-PIC solution, it is possible to benefit from all its benefits (most notably convergence that does not depend in the number of prefixes) without any requirement for other routers to support BGP-PIC.Bashandy Expires January 7, 2025 [Page 19]
Internet-Draft BGP Prefix Independent Convergence July 20248. Security Considerations The behavior described in this document is internal functionality to a router that result in significant improvement to convergence time as well as reduction in CPU and memory used by FIB while not showing change in basic routing and forwarding functionality. As such no additional security risk is introduced by using the mechanisms described in this document.9. IANA Considerations This document has no IANA actions.10. References10.1. Normative References [RFC4271] Rekhter, Y., Li, T., and S. Hares, "A Border Gateway Protocol 4 (BGP-4),RFC 4271, January 2006. [RFC3031] E. Rosen, A. Viswanathan, R. Callon, "Multiprotocol Label Switching Architecture",RFC 3031, January 200110.2. Informative References [I-D.ietf-idr-best-external] Marques,P., Fernando, R., Chen, E, Mohapatra, P., Gredler, H., "Advertisement of the best external route in BGP",draft-ietf-idr-best-external-05.txt, January 2012. [RFC5565] Wu, J., Cui, Y., Metz, C., and E. Rosen, "Softwire Mesh Framework",RFC 5565, June 2009. [RFC4364] Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private Networks (VPNs)",RFC 4364, February 2006. [RFC4798] De Clercq, J. , Ooms, D., Prevost, S., Le Faucheur, F., "Connecting IPv6 Islands over IPv4 MPLS Using IPv6 Provider Edge Routers (6PE)",RFC 4798, February 2007. [bonaventure] O. Bonaventure, C. Filsfils, and P. Francois. "Achieving sub-50 milliseconds recovery upon bgp peering link failures, " IEEE/ACM Transactions on Networking, 15(5):1123-1135, 2007Bashandy Expires January 7, 2025 [Page 20]
Internet-Draft BGP Prefix Independent Convergence July 2024 [RFC5036] Andersson, L., Minei, I., and B. Thomas, "LDP Specification", RFC, October 2007 [RFC7911] D. Walton, A. Retana, E. Chen, J. Scudder, "Advertisement of Multiple Paths in BGP",RFC 7911, July 2016 [RFC6774] R. Raszuk, R. Fernando, K. Patel, D. McPherson, K. Kumaki, "Distribution of diverse BGP paths",RFC 6774, November 2012 [I-D.pmohapat-idr-fast-conn-restore] P. Mohapatra, R. Fernando, C. Filsfils, and R. Raszuk, "Fast Connectivity Restoration Using BGP Add-path",draft-pmohapat-idr-fast-conn-restore-03, Jan 2013 [I-D.ietf-rtgwg-segment-routing-ti-lfa] S. Litkowski, A. Bashandy, C. Filsfils, P. Francois, B. Decraene, D. Voyer, "Topology Independent Fast Reroute using Segment Routing",draft-ietf-rtgwg-segment-routing-ti-lfa-09 (work in progress), December 2022 [RFC5714] M. Shand and S. Bryant, "IP Fast Reroute Framework",RFC5714, January 2010 [RFC7490] S. Bryant, C. Filsfils, S. Previdi, M. Shand, N So, " Remote Loop-Free Alternate (LFA) Fast Reroute (FRR)",RFC7490 April 2015 [RFC7812] A. Atlas, C. Bowers, G. Enyedi, " An Architecture for IP/LDP Fast-Reroute Using Maximally Redundant Trees",RFC7812, June 2016 [RFC8277] E. Rosen, " Carrying Label Information in BGP-4",RFC8277, October 2017 [RFC8660] A. Bashandy, C. Filsfils, S. Previdi, B. Decraene, S. Litkowski, M. Horneffer, R. Shakir, "Segment Routing with MPLS data plane",RFC 8660, December 2019 [RFC9107] R. Raszuk, B. Decraene, C. Cassar, E. Aman, K Wang, " BGP Optimal Route Reflection (BGP ORR)",RFC9107, August 202111. Acknowledgments Special thanks to Neeraj Malhotra and Yuri Tsier for the valuable helpBashandy Expires January 7, 2025 [Page 21]
Internet-Draft BGP Prefix Independent Convergence July 2024 Special thanks to Bruno Decraene, Theresa Enghardt, Ines Robles, Luc Andre Burdet, and Alvaro Retana for the valuable comments This document was prepared using 2-Word-v2.0.template.dot.Authors' Addresses Ahmed Bashandy Cisco Systems Email: abashandy.ietf@gmail.com Clarence Filsfils Cisco Systems Brussels, Belgium Email: cfilsfil@cisco.com Prodosh Mohapatra Sproute Networks Email: mpradosh@yahoo.comBashandy Expires January 7, 2025 [Page 22]
Internet-Draft BGP Prefix Independent Convergence July 2024Appendix A. Handling Platforms with Limited Levels of Hierarchy This section provides additional details on how to handle platforms with limited number of hierarchical levels. Let's consider a pathlist associated with the leaf "R1" consisting of the list of pic-paths <P1, P2,..., Pn>. Assume that the leaf "R1" has an OutLabel-list <L1, L2,..., Ln>. Suppose the pic-path Pi is a recursive pic-path that resolves via a prefix represented by the leaf "R2". The leaf "R2" itself is pointing to a pathlist consisting of the pic-paths <Q1, Q2,..., Qm>. If the platform supports the number of hierarchy levels of the forwarding chain, then a packet that uses the pic-path "Pi" will be forwarded according to the steps inSection 4. Suppose the platform cannot support the number of hierarchy levels in the forwarding chain. FIB manager needs to reduce the number of hierarchy levels when programming the forwarding chain in the FIB. The idea of reducing the number of hierarchy levels is to "flatten" two chain levels into a single level. The "flattening" steps are as follows 1. FIB manager walks to the parent of "Pi", which is the leaf "R2". 2. FIB manager extracts the parent pathlist of the leaf "R2", which is <Q1, Q2,..., Qm>. 3. FIB manager also extracts the OutLabel-list of R2 associated with the leaf "R2". Remember that the OutLabel-list of R2 is <L1, L2,..., Lm>. 4. FIB manager replaces the pic-path "Pi", with the list of pic- paths <Q1, Q2,..., Qm>. 5. Hence the pic-path list <P1, P2,..., Pn> now becomes "<P1, P2,...,Pi-1, Q1, Q2,..., Qm, Pi+1, Pn>. 6. The path-index stored inside the locations "Q1", "Q2", ..., "Qm" must all be "i" because the index "i" refers to the label "Li" associated with leaf "R1". 7. FIB manager attaches an OutLabel-list with the new pathlist as follows: <Unlabeled,..., Unlabeled, L1, L2,..., Lm, Unlabeled, ..., Unlabeled>. The size of the label list associated with the flattened pathlist equals the size of the pathlist. Thus there is a 1-1 mapping between every pic-path in the "flattened" pathlist and the OutLabel-list associated with it.Bashandy Expires January 7, 2025 [Page 23]
Internet-Draft BGP Prefix Independent Convergence July 2024 It is noteworthy to mention that the labels in the OutLabel-list associated with the "flattened" pathlist may be stored in the same memory location as the pic-path itself to avoid additional memory access. The same steps can be applied to all pic-paths in the pathlist <P1, P2,..., Pn> so that all pic-paths are "flattened" thereby reducing the number of hierarchical levels by one. Note that that "flattening" a pathlist pulls in all pic-paths of the parent pic- paths, a desired feature to utilize all pic-paths at all levels. A platform that has a limit on the number of pic-paths in a pathlist for any given leaf may choose to reduce the number pic-paths using methods that are beyond the scope of this document. The steps can be recursively applied to other pic-paths at the same levels or other levels to recursively reduce the number of hierarchical levels to an arbitrary value so as to accommodate the capability of the forwarding engine. Because a flattened pathlist may have an associated OutLabel-list the forwarding behavior has to be slightly modified. The modification is done by adding the following step right after step 4 inSection 4. 5. If there is an OutLabel-list associated with the pathlist, then if the pic-path "Pi" is chosen by the hashing algorithm, retrieve the label at location "i" in that OutLabel-list and apply the label action of that label on the packet. The steps in this Section to are applied to an example in the next Section.Bashandy Expires January 7, 2025 [Page 24]
Internet-Draft BGP Prefix Independent Convergence July 2024Appendix B. Example: Flattening a forwarding chain. This example uses a case of inter-AS option C [RFC4364] where there are 3 levels of hierarchy. Figure 4 illustrates the sample topology. The Autonomous System Border Routers (ASBRs) on the ingress domain (Domain 1) use BGP to advertise the core routers (ASBRs and ePEs) of the egress domain (Domain 2) to the iPE. The end result is that the ingress PE (iPE) has 2 levels of recursion for the VPN prefixes VPN- IP1 and VPN-IP2. Domain 1 Domain 2 +-------------+ +-------------+ | | | | | LDP/SR Core | | LDP/SR core | | | | | | (192.0.2.4) | | | ASBR11-------ASBR21........ePE1(192.0.2.1) | | \ / | . . |\ | | \ / | . . | \ | | \ / | . . | \ | | \/ | .. | \VPN-IP1(198.51.100.0/24) | | /\ | . . | /VRF "Blue" ASN: 65000 | | / \ | . . | / | | / \ | . . | / | | / \ | . . |/ iPE ASBR12-------ASBR22........ePE2 (192.0.2.2) | (192.0.2.5) | |\ | | | | \ | | | | \ | | | | \VRF "Blue" ASN: 65000 | | | | /VPN-IP2(203.0.113.0/24) | | | | / | | | | / | | | |/ | ASBR13-------ASBR23........ePE3(192.0.2.3) | (192.0.2.6) | | | | | | | | | | +-------------+ +-------------+ <=========== <========= <============ Advertise ePEx Advertise Redistribute Using IBGP-LU ePEx Using ePEx routes EBGP-LU into BGP Figure 4: Sample 3-level hierarchy topologyBashandy Expires January 7, 2025 [Page 25]
Internet-Draft BGP Prefix Independent Convergence July 2024 The following assumptions about connectivity are made: o In "Domain 2", both ASBR21 and ASBR22 can reach both ePE1 and ePE2 using the same metric. o In "Domain 2", only ASBR23 can reach ePE3. o In "Domain 1", iPE (the ingress PE) can reach ASBR11, ASBR12, and ASBR13 via IGP using the same metric. The following assumptions are made about the labels: o The VPN labels advertised by ePE1 and ePE2 for prefix VPN-IP1 are VPN-L11 and VPN-L21, respectively. o The VPN labels advertised by ePE2 and ePE3 for prefix VPN-IP2 are VPN-L22 and VPN-L32, respectively. o The labels advertised by ASBR11 to iPE using BGP-LU for the egress PEs ePE1 and ePE2 are LASBR111(ePE1) and LASBR112(ePE2), respectively. o The labels advertised by ASBR12 to iPE using BGP-LU for the egress PEs ePE1 and ePE2 are LASBR121(ePE1) and LASBR122(ePE2), respectively. o The label advertised by ASBR13 to iPE using BGP-LU for the egress PE ePE3 is LASBR13(ePE3). o The IGP labels advertised by the next hops directly connected to iPE towards ASBR11, ASBR12, and ASBR13 in the core of domain 1 are IGP-L11, IGP-L12, and IGP-L13, respectively. o Both the routers ASBR21 and ASBR22 of Domain 2 advertise the same label LASBR21 and LASBR22 for the egress PEs ePE1 and ePE2, respectively, to the routers ASBR11 and ASBR22 of Domain 1. o The router ASBR23 of Domain 2 advertises the label LASBR23 for the egress PE ePE3 to the router ASBR13 of Domain 1. Based on these connectivity assumptions and the topology in Figure 4, the routing table on iPE isBashandy Expires January 7, 2025 [Page 26]
Internet-Draft BGP Prefix Independent Convergence July 2024 65000: 198.51.100.0/24 via ePE1 (192.0.2.1), VPN Label: VPN-L11 via ePE2 (192.0.2.2), VPN Label: VPN-L21 65000: 203.0.113.0/24 via ePE2 (192.0.2.2), VPN Label: VPN-L22 via ePE3 (192.0.2.3), VPN Label: VPN-L32 192.0.2.1/32 (ePE1) via ASBR11, BGP-LU Label: LASBR111(ePE1) via ASBR12, BGP-LU Label: LASBR121(ePE1) 192.0.2.2/32 (ePE2) via ASBR11, BGP-LU Label: LASBR112(ePE2) via ASBR12, BGP-LU Label: LASBR122(ePE2) 192.0.2.3/32 (ePE3) Via ASBR13, BGP-LU Label: LASBR13(ePE3) 192.0.2.4/32 (ASBR11) via Core, Label: IGP-L11 192.0.2.5/32 (ASBR12) via Core, Label: IGP-L12 192.0.2.6/32 (ASBR13) via Core, Label: IGP-L13 The diagram in Figure 5 illustrates the forwarding chain in iPE assuming that the forwarding hardware in iPE supports 3 levels of hierarchy. The leaves corresponding to the ASBRs on domain 1 (ASBR11, ASBR12, and ASBR13) are at the bottom of the hierarchy. There are few important points: o Because the hardware supports the required depth of hierarchy, the sizes of a pathlist equal the size of the label list associated with the leaves using this pathlist. o The path-index inside the pathlist entry indicates the label that will be picked from the OutLabel-List associated with the child leaf if that pic-path is chosen by the forwarding engine hashing function.Bashandy Expires January 7, 2025 [Page 27]
Internet-Draft BGP Prefix Independent Convergence July 2024 OutLabel-List OutLabel-List For VPN-IP1 For VPN-IP2 +------------+ +--------+ +-------+ +------------+ | VPN-L11 |<---| VPN-IP1| |VPN-IP2|-->| VPN-L22 | +------------+ +---+----+ +---+---+ +------------+ | VPN-L21 | | | | VPN-L32 | +------------+ | | +------------+ | | V V +---+---+ +---+---+ | 0 | 1 | | 0 | 1 | +-|-+-\-+ +-/-+-\-+ | \ / \ | \ / \ | \ / \ | \ / \ v \ / \ +-----+ +-----+ +-----+ +----+ ePE1| |ePE2 +-----+ | ePE3+-----+ | +--+--+ +-----+ | +--+--+ | v | / v | v +--------------+ | / +--------------+ | +-------------+ |LASBR111(ePE1)| | / |LASBR112(ePE2)| | |LASBR13(ePE3)| +--------------+ | / +--------------+ | +-------------+ |LASBR121(ePE1)| | / |LASBR122(ePE2)| | OutLabel-List +--------------+ | / +--------------+ | For ePE3 OutLabel-List | / OutLabel-List | For ePE1 | / For ePE2 | | / | | / | | / | v v v +---+---+ Shared pathlist +---+ pathlist | 0 | 1 | For ePE1 and ePE2 | 0 | For ePE3 +-|-+-\-+ +-|-+ | \ | | \ | | \ | | \ | v v v +------+ +------+ +------+ +---+ASBR11| |ASBR12+--+ |ASBR13+---+ | +------+ +------+ | +------+ | v v v +-------+ +-------+ +-------+ |IGP-L11| |IGP-L12| |IGP-L13| +-------+ +-------+ +-------+ Figure 5: Forwarding Chain for hardware supporting 3 LevelsBashandy Expires January 7, 2025 [Page 28]
Internet-Draft BGP Prefix Independent Convergence July 2024 Now suppose the hardware on iPE (the ingress PE) supports 2 levels of hierarchy only. In that case, the 3-levels forwarding chain in Figure 5 needs to be "flattened" into 2 levels only. OutLabel-List OutLabel-List For VPN-IP1 For VPN-IP2 +------------+ +-------+ +-------+ +------------+ | VPN-L11 |<---|VPN-IP1| | VPN-IP2|--->| VPN-L22 | +------------+ +---+---+ +---+---+ +------------+ | VPN-L21 | | | | VPN-L32 | +------------+ | | +------------+ | | | | | | Flattened | | Flattened pathlist V V pathlist +===+===+ +===+===+===+ +==============+ +--------+ 0 | 1 | | 0 | 0 | 1 +---->|LASBR112(ePE2)| | +=|=+=\=+ +=/=+=/=+=\=+ +==============+ v | \ / / \ |LASBR122(ePE2)| +==============+ | \ +-----+ / \ +==============+ |LASBR111(ePE1)| | \/ / \ |LASBR13(ePE3) | +==============+ | /\ / \ +==============+ |LASBR121(ePE1)| | / \ / \ +==============+ | / \ / \ | / \ / \ | / + + \ | + | | \ | | | | \ v v v v v +------+ +------+ +------+ +----|ASBR11| |ASBR12+---+ |ASBR13+---+ | +------+ +------+ | +------+ | v v v +-------+ +-------+ +-------+ |IGP-L11| |IGP-L12| |IGP-L13| +-------+ +-------+ +-------+ Figure 6: Flattening 3 levels to 2 levels of Hierarchy on iPE Figure 6 represents one way to "flatten" a 3 levels hierarchy into two levels. There are a few important points:Bashandy Expires January 7, 2025 [Page 29]
Internet-Draft BGP Prefix Independent Convergence July 2024 o As mentioned in SectionAppendix A, a flattened pathlist may have label lists associated with them. The size of the label list associated with a flattened pathlist equals the size of the pathlist. Hence it is possible that an implementation includes these label lists in the flattened pathlist itself. o Again as mentioned in SectionAppendix A, the size of a flattened pathlist may not be equal to the size of the OutLabel-lists of leaves using the flattened pathlist. So the indices inside a flattened pathlist still indicate the label index in the OutLabel-Lists of the leaves using that pathlist. Because the size of the flattened pathlist may be different from the size of the OutLabel-lists of the leaves, the indices may be repeated. o Let's take a look at the flattened pathlist used by the prefix "VPN-IP2". The pathlist associated with the prefix "VPN-IP2" has three entries. o The first and second entry have index "0". This is because both entries correspond to ePE2. Thus when hashing performed by the forwarding engine results in using the first or the second entry in the pathlist, the forwarding engine will pick the correct VPN label "VPN-L22", which is the label advertised by ePE2 for the prefix "VPN-IP2". o The third entry has the index "1". This is because the third entry corresponds to ePE3. Thus when the hashing is performed by the forwarding engine results in using the third entry in the flattened pathlist, the forwarding engine will pick the correct VPN label "VPN-L32", which is the label advertised by "ePE3" for the prefix "VPN-IP2". Now let's try and apply the forwarding steps inSection 4 together with the additional step in SectionAppendix A to the flattened forwarding chain illustrated in Figure 6. o Suppose a packet arrives at "iPE" and matches the VPN prefix "VPN-IP2". o The forwarding engine walks to the parent of the "VPN-IP2", which is the flattened pathlist and applies a hashing algorithm to pick a pic-path. o Suppose the hashing by the forwarding engine picks the second pic-path in the flattened pathlist associated with the leaf "VPN- IP2". o Because the second pic-path has the index "0", the label "VPN- L22" is pushed on the packet.Bashandy Expires January 7, 2025 [Page 30]
Internet-Draft BGP Prefix Independent Convergence July 2024 o Next the forwarding engine picks the second label from the OutLabel-List associated with the flattened pathlist resulting in "LASBR122(ePE2)" being the next pushed label. o The forwarding engine now moves to the parent of the flattened pathlist corresponding to the second pic-path. The parent is the IGP label leaf corresponding to "ASBR12". o So the packet is forwarded towards the ASBR "ASBR12" and the IGP label at the top will be "IGP-L12". Based on the above steps, a packet arriving at iPE and destined to the prefix VPN-L22 reaches its destination as follows: o iPE sends the packet along the shortest pic-path towards ASBR12 with the following label stack starting from the top: {L12, LASBR122(ePE2), VPN-L22}. o The penultimate hop of ASBR12 pops the top label "L12". Hence the packet arrives at ASBR12 with the remaining label stack {LASBR122(ePE2), VPN-L22} where "LASBR12(ePE2)" is the top label. o ASBR12 swaps "LASBR122(ePE2)" with the label "LASBR22(ePE2)", which is the label advertised by ASBR22 for the ePE2 (the egress PE). o ASBR22 receives the packet with "LASBR22(ePE2)" at the top. o Hence ASBR22 swaps "LASBR22(ePE2)" with the IGP label for ePE2 advertised by the next-hop towards ePE2 in domain 2, and sends the packet along the shortest pic-path towards ePE2. o The penultimate hop of ePE2 pops the top label. Hence ePE2 receives the packet with the top label VPN-L22 at the top o ePE2 pops "VPN-L22" and sends the packet as a pure IP packet towards the destination VPN-IP2.Bashandy Expires January 7, 2025 [Page 31]
Internet-Draft BGP Prefix Independent Convergence July 2024Appendix C. Perspective The following table puts the BGP-PIC benefits in perspective assuming o 1M impacted BGP prefixes o IGP convergence ~ 500 msec o local protection ~ 50msec o FIB Update per BGP destination ~ 100usec conservative, ~ 10usec optimistic o BGP best route recalculation per BGP destination ~ 10usec optimistic, ~ 100usec optimistic Without PIC With PIC Local IGP Failure 10 to 100sec 50msec Local BGP Failure 100 to 200sec 50msec Remote IGP Failure 10 to 100sec 500msec Local BGP Failure 100 to 200sec 500msec Upon local IGP next-hop failure or remote IGP next-hop failure, the existing primary BGP next-hop is intact and usable hence the resiliency only depends on the ability of the FIB mechanism to reflect the new pic-path to the BGP next-hop to the depending BGP destinations. Without BGP-PIC, a conservative back-of-the-envelope estimation for this FIB update is 100usec per BGP destination. An optimistic estimation is 10usec per entry. Upon local BGP next-hop failure or remote BGP next-hop failure, without the BGP-PIC mechanism, a new BGP Best-Path needs to be recomputed and new updates need to be sent to peers. This depends on BGP processing time that will be shared between best-path computation, RIB update and peer update. A conservative back-of-the- envelope estimation for this is 200usec per BGP destination. An optimistic estimation is 100usec per entry.Bashandy Expires January 7, 2025 [Page 32]
Datatracker
draft-ietf-rtgwg-bgp-pic-21
This is an older version of an Internet-Draft whose latest revision state is "Active".
| Document | Document type | This is an older version of an Internet-Draft whose latest revision state is "Active". Expired & archived | |
|---|---|---|---|
| Select version | |||
| Compare versions | |||
| Authors | Ahmed Bashandy,Clarence Filsfils,Prodosh Mohapatra | ||
| Replaces | draft-bashandy-rtgwg-bgp-pic | ||
| RFC stream |   | ||
| Other formats | |||
| Additional resources | Mailing list discussion |