| Internet-Draft | NNHN | November 2026 |
| Wang, et al. | Expires 20 May 2027 | [Page] |
BGP speakers learn their next-hop addresses for NLRI in RFC 4271 in the NEXT_HOP field and in RFC 4760 in the "Network Address of Next Hop" field. Under certain circumstances, it might be desirable for a BGP speaker to know both the next-hops and the next-next-hops of NLRI to make optimal forwarding decisions. One such example is global load balancing (GLB) in a Clos network.¶
Draft-ietf-idr-nhc defines the "Next Hop Dependent Characteristics Attribute" (NHC) which allows a BGP speaker to signal the forwarding characteristics associated with a given next-hop.¶
This document defines a new NHC characteristic, the Next-next Hop Nodes (NNHN) characteristic, which can be used to advertise the next-next-hop nodes associated with a given next-hop.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
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."¶
This Internet-Draft will expire on 20 May 2027.¶
Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
BGP speakers learn their next-hop addresses for NLRI in [RFC4271] in the NEXT_HOP field and in [RFC4760] in the "Network Address of Next Hop" field. Under certain circumstances, it might be desirable for a BGP speaker to know both the next-hops and the next-next-hops of NLRI to make optimal forwarding decisions. One such example is the global load balancing (GLB) in a Clos network [I-D.cheng-rtgwg-adaptive-routing-framework].¶
When a route's ECMP has multiple next-hops, packets forwarded using that ECMP are hashed to the member next-hops for load balancing purposes. If one of the member next-hop links is congested due to uneven hashing, dynamic load balancing (DLB) allows the node to adjust the hashing so that the congestion on that link can be mitigated. When all next-hop link(s) are congested, DLB on the local node will not help to mitigate the congestion. Such nodes will require help from the previous hop to shift the traffic towards alternative nodes to mitigate such congestion. This process is called global load balancing.¶
In a Clos network, a congested link will affect the load balancing decisions of the previous layer nodes equally. Because of this, the previous-previous layer nodes do not need to change their load balancing decisions towards the previous layer nodes to mitigate this link congestion. This means we only need to know the link congestion status of the next-next-hops of given BGP route in order to make GLB decisions. The combined link quality (congestion/failure) of each next-hop and its corresponding next-next-hops can be used as the feedback for DLB.¶
The purpose of this document is to provide a method for BGP to learn the next-next-hops, or more specifically, the next-next-hop nodes. When a next-hop node has parallel links towards a next-next-hop node, DLB helps to balance the load between the parallel links by locally adjusting the volume of traffic hashed over a given ECMP member link. Thus, only the overall link congestion between the next-hop node and the next-next-hop node is important for GLB.¶
Note that the mechanism for detecting link congestion and communicating them to the previous hop nodes is out of the scope of this document. [I-D.zzhang-fann-router-info] defined one approach to notify link quality to directly connected nodes.¶
This document defines a new NHC characteristic, the Next-next Hop Nodes (NNHN) characteristic, for the BGP Next Hop Dependent Characteristics Attribute (NHC) defined in [I-D.ietf-idr-nhc]. A downstream BGP speaker can use the NNHN to advertise the next-next-hop nodes corresponding to the next-hop of an NLRI. This allows the upstream BGP speaker to learn the next-next-hop nodes corresponding to each of its next-hop nodes.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
[I-D.ietf-idr-nhc] defines NHC as a container for characteristic TLVs. Next-next Hop Nodes is one such characteristic. It specifies the next-next-hop nodes corresponding to the next-hop field in the NHC.¶
The NNHN TLV has the NHC characteristic code 2, as assigned in Section 4 of [I-D.ietf-idr-nhc]. The NHC characteristic length is variable. The NNHN characteristic format is shown in Figure 1:¶
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Characteristic Code = 2 |Characteristic Length(variable)| +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Next-hop BGP ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ Next-next-hop BGP IDs (variable) ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
All procedures from Section 2.2 of [I-D.ietf-idr-nhc] apply.¶
When a BGP speaker S has a BGP route R it wishes to advertise with next-hop self to its peer, it MAY choose to originate an NNHN characteristic. The "Next-hop BGP ID" field MUST be set to the BGP Identifier this BGP speaker uses with the peer.¶
If route R has only a single path, the BGP Identifier of the BGP peer that advertised the path MUST be encoded in the "Next-next-hop BGP IDs" field.¶
If route R has multiple paths, for all the ECMP paths which are used for forwarding, the BGP Identifiers of those BGP peers that advertised the ECMP paths MUST be encoded in the "Next-next-hop BGP IDs" field.¶
When there are more than one "Next-next-hop BGP IDs" in the characteristic, they MUST be encoded in the numerically ascending order, with no duplicates, treating the BGP Identifier as a network byte order encoded 32-bit unsigned integer.¶
An NNHN MUST have at least one "Next-next-hop BGP ID".¶
When a BGP speaker S has a BGP route R it wishes to advertise with next-hop self to its peer, it MUST NOT forward the NNHN characteristic received from downstream peers. It either originates its own NNHN characteristic as described above or does not send one.¶
When a BGP speaker S has a BGP route R it wishes to advertise with the next-hop that has not been set to self, it MUST NOT originate an NNHN characteristic. However, if a NNHN characteristic has been received for route R and passed the NHC validation as defined in [I-D.ietf-idr-nhc], the NNHN characteristic SHOULD be forwarded.¶
A BGP speaker MUST NOT include more than one instance of NNHN in an NHC.¶
When the NNHN is originated locally due to next-hop self, there is no case in which it needs to be aggregated. When the NNHN is not originated locally because the next-hop is not set to self, we MUST NOT forward the NNHN unless all constituent routes of the aggregate route have the same NNHN.¶
All procedures from Section 2.3 of [I-D.ietf-idr-nhc] apply.¶
The receiver of the NNHN characteristic MUST be able to handle any order of the "Next-next-hop BGP IDs".¶
The receiver of the NNHN characteristic uses the "Next-hop BGP ID" and "Next-next-hop BGP IDs" fields to determine the set of next-next-hops used by the next-hop of this route. Usage of this information is beyond the scope of this document. Please refer to Section 1 for some of the usage examples.¶
The NNHN characteristic length MUST be at least 8 and MUST be a multiple of 4, otherwise it is malformed. Malformed NNHN characteristics MUST be discarded.¶
If more than one instance of NNHN is included in an NHC, instances beyond the first MUST be discarded.¶
When an NNHN characteristic is received from a single-hop EBGP peer and the "Next-hop BGP Identifier" does not match the peer's BGP Identifier, it MUST be discarded.¶
When duplicate "Next-next-hop BGP IDs" exist in an NNHN, only the first one is accepted and the remaining duplicates MUST be discarded.¶
NNHN characteristic is intended to be used under data center Clos network described in [RFC7938]. Since BGP Identifiers are used to identify the next-next-hop nodes, an operation MUST ensure that the BGP Identifiers are unique across the Clos network where NNHN characteristic is sent.¶
In the event of a topology change, such as a link failure, propagating the NNHN characteristic may increase the number of BGP route updates. For example, in a five-stage Clos network, if a link between a spine and a leaf fails, the spine must withdraw the leaf route from the super spines. However, if a super spine advertises the NNHN characteristic for the leaf route, it must also send route updates to the other spines to reflect the change in NNHN, since the leaf route on the super spine has lost one of its ECMP next-hop legs. These additional route updates sent by the super spine to the other spines represent the overhead introduced by NNHN propagation. A typical super spine in a Clos network may connect to 32 spines. Considering that the number of routes received from any individual leaf is usually limited and that link failures within the network are relatively infrequent, the additional route updates resulting from NNHN propagation should be manageable.¶
Since NHC Characteristic Code 2, has already been assigned in Section 5 of [I-D.ietf-idr-nhc], for the NNHN characteristic, there is no further IANA actions for this document.¶
[The RFC-Editor should remove this section before publishing this document as an RFC]¶
Global load balancing using NNHN characteristic has been implemented by HPE Networking Juniper and New H3C Technologies. Interoperation tests between both implementations were successfully conducted in 2024.¶
As NNHN is not cryptographically protected, modification or insertion of a syntactically valid but bogus NNHN characteristic by an attacker would perturbate the applications using the knowledge of these next-next-hops.¶
The authors would like to thank John Scudder, Jie Dong, Robert Raszuk, Jordan Head, Andrew Stone, Gyan Mishra for their reviews, comments, and suggestions. Special thanks to Bruno Decraene for his extremely thoughtful review and even writeup suggestions for some of the paragraphs.¶