Traffic Steering using BGP FlowSpec with SR Policy

SR-MPLS forwards data packets using the source routing model. The core idea of SR-MPLS is to divide a packet forwarding path into different segments, allocate segment identifiers (SIDs) to the segments, and encapsulate segment information into packets at the ingress of the path to guide packet forwarding. Segment Routing IPv6 (SRv6) is a protocol designed to forward IPv6 data packets on a network using the source routing model. SRv6 enables the ingress network device to add a segment routing header (SRH) to an IPv6 packet and push an explicit IPv6 address stack into the SRH. After receiving the packet, each transit node updates the IPv6 destination IP address in the packet and segment list to implement hop-by-hop forwarding. SR Policy (includes SR-MPLS and SRv6 Policy) is a tunneling technology based on SR-MPLS and SRv6. An SR Policy is a set of candidate paths consisting of one or more segment lists, that is, segment ID (SID) lists. Each SID list identifies an end-to-end path from the source node to the destination node, instructing a network device to forward traffic through the path rather than the shortest path computed using an IGP. The header of a packet steered into an SR Policy is augmented with an ordered list of segments associated with that SR Policy, so that other devices on the network can execute the instructions encapsulated into the list. The headend of an SR Policy may learn multiple candidate paths for an SR Policy. Candidate paths may be learned via a number of different mechanisms, e.g., CLI, NetConf, PCEP , or BGP . , and define the BGP Flow Specification that allows conveying Flow Specifications and traffic Action/Rules associated (rate- limiting, redirect, remark ...). BGP Flow Specifications are encoded within the MP_REACH_NLRI and MP_UNREACH_NLRI attributes . Rules (Actions associated) are encoded in Extended Community attribute . The BGP Flow Specification function allows BGP Flow Specification routes that carry traffic policies to be transmitted to BGP Flow Specification peers to steer traffic. This document specifies the Extensions to BGP Flow Specification (FlowSpec) to steer matching traffic flows into a Segment Routing (SR) Policy . Specifically, it defines the normative protocol procedures for combining the FlowSpec Network Layer Reachability Information (NLRI) with the BGP Prefix-SID Attribute [RFC8669] and specific BGP Extended Communities to signal SR-MPLS and SRv6 Policy steering pathways. This document does not modify the base Segment Routing architecture [RFC8402] or SR Policy constructs [RFC9256], but rather establishes the interworking protocol semantics between BGP FlowSpec controllers and SR Headends on the Standards Track.

FlowSpec: Flow Specification SR: Segment Routing SR-MPLS: SR over the MPLS data plane SRv6: SR over the IPv6 data plane SID: Segment Identifier SRH: Segment Routing Header USD: Ultimate Segment Decapsulation

Per , the usage of the BGP Prefix-SID attribute for AFI/SAFI combinations other than Labeled Unicast was left to future specifications. This document serves as that explicit specification for the FlowSpec AFI/SAFI, without modifying the attribute's core structural semantics defined in .

When a BGP FlowSpec route is used to steer traffic into an SR Policy, the BGP FlowSpec NLRI defines the traffic classification criteria. The steering target (the SR Policy) is identified by the combination of the following attributes attached to the FlowSpec route: BGP Prefix-SID Attribute : Used to signal the Segment Routing Identifiers.The BGP Prefix-SID defined in is utilized to enable SRv6 VPN services [RFC9252]. SRv6 Services TLVs within the BGP Prefix-SID Attribute can be used to indicate the endpoint functions. This document extends the use of the BGP Prefix-SID attribute to carry SRv6 SIDs and their associated information with the BGP address families that are defined in and , where applicable, as described in Section 5. Color Extended Community : Carries the 4-octet Color value of the target SR Policy, providing the steering intent. Redirect-to-IP Extended Community (or Redirect-to-IPv6): Carries the IP address of the target SR Policy's Endpoint. For the case where a FlowSpec route carries multiple Color Extended Communities, the Color Extended community with the highest numerical value MUST be given higher preference to resolve the target SR Policy, per the description in Section 8.4.1 of .

A BGP Speaker or a centralized Controller acting as a BGP FlowSpec sender MUST adhere to the following procedures when constructing and advertising an SR Policy steering FlowSpec route: Traffic Identification: The sender MUST encode the targeted flow characteristics into the FlowSpec NLRI following and . Steering Resolution: The sender MUST attach the Color Extended Community to specify the Policy Color, and a Redirect Extended Community to specify the Policy Endpoint. Path/Segment Signaling: To enforce specific SR node/link pathways or segment containment, the sender MUST attach the BGP Prefix-SID Attribute . When targeting an SRv6 infrastructure, the appropriate SRv6 Service TLV and SRv6 SID MUST be fully populated according to the encoding rules defined in Section 4.2 of . Propagation: The sender MUST advertise the resulting BGP UPDATE message within the FlowSpec AFI/SAFI (AFI=1/2, SAFI=133/134). For SRv6 scenarios, to support Ultimate Segment Decapsulation (USD) flavors at the tailend, the sender behaviors MUST satisfy the following conditions: If the last SID in the SR segment list programmed within the target SRv6 Policy is a USD-flavored SRv6 SID, attaching the BGP Prefix-SID Attribute to convey the USD flavor is OPTIONAL. Otherwise, the BGP Prefix-SID attribute MUST be carried by the sender to explicitly convey a USD-flavored SRv6 SID to indicate specific Tailend decapsulation functions.

A BGP FlowSpec receiver (typically an SR Headend or Tailend device) processing a FlowSpec route with SR Policy attributes MUST execute the following validation and installation procedures: Validation: The receiver MUST first validate the FlowSpec NLRI against the base rules defined in /. If validation fails, error handling according to and applies to this specification. Steering Matching: The receiver MUST extract the Color, Endpoint, and Prefix-SID from the path attributes. It MUST attempt to resolve these parameters against its local SR Policy Database . Resolution Failure Handling: Route Retention (BGP Level): If the Color, Endpoint, or Prefix-SID cannot be resolved to an active/valid SR Policy in the local SR Database, the receiver MUST NOT discard or withdraw the BGP FlowSpec route itself. The BGP FlowSpec route MUST be retained in the BGP RIB (Loc-RIB). Data Plane Steering Fallback: The steering action to the unresolved SR Policy MUST NOT be installed into the local Forwarding Information Base (FIB). Action Execution: If other valid FlowSpec traffic filtering actions (e.g., rate-limit, traffic-remarking, or drop) are present in the same FlowSpec route, the receiver MUST continue to apply those filtering actions in the data plane. For the packet matching the FlowSpec NLRI, the receiver MUST fall back to default IP/SRv6 forwarding (native shortest-path forwarding) rather than steering into the unresolved SR Policy. State Re-evaluation: The receiver MUST monitor the local SR Policy Database. Once the corresponding SR Policy becomes active/valid, the receiver MUST dynamically update its FIB to activate the SR Policy steering action for the matching FlowSpec traffic without requiring a BGP route refresh. Forwarding Programming: If a valid local SR Policy matches the extracted <Color, Endpoint>, the receiver MUST program the hardware forwarding plane to encapsulate the matching FlowSpec traffic into the associated SR Policy segment list. Handling Programming Failure: If the forwarding entry cannot be programmed into the hardware forwarding plane (e.g., due to TCAM/FIB resource exhaustion, unsupported hardware combination, or platform-specific limits): Error Logging: The receiver MUST generate a system log (Syslog) or local management notification (e.g., SNMP trap or YANG notification) indicating the programming failure, the associated FlowSpec NLRI, and the specific reason for failure. Data Plane Traffic Handling: The receiver MUST NOT drop traffic matching the FlowSpec NLRI solely due to a programming failure, unless a explicit filtering action (e.g., Traffic-Rate 0 / Drop) is mandated by the same FlowSpec route. Instead, traffic MUST fall back to standard IP/SRv6 shortest-path forwarding (best-effort forwarding) or follow local policy configurations. Resource Monitoring and Retry: The receiver SHOULD monitor data plane hardware resources. Once hardware resources become available, the receiver MUST automatically attempt to re-program the un-installed FlowSpec forwarding entries. The method specified in this document supports load balancing to the tailend device. To achieve this, the Headend device MAY set up multiple candidate paths within the resolved SR Policy, and use the matching FlowSpec route to steer traffic into that multi-path SR Policy.

This specification introduces zero modifications to the existing BGP FlowSpec validation state-machines. If a receiving device supports base FlowSpec [RFC8955] but does not recognize the BGP Prefix-SID attribute in the context of FlowSpec SAFI, it MUST follow the standard BGP transitive attribute handling rules. It will safely ignore the Prefix-SID attribute and fall back to evaluating the Redirect Extended Community, attempting a standard IP/IPv6 redirection. This ensures graceful backward compatibility.

The error handling procedures depend on the results of the following: a) The validation procedures of the redirect-to-IP Extended Community as per b) The validation and selection procedures of the Color Extended Community as per , which determine a single color for steering c) The validation procedures of the Prefix-SID attribute per section 6 of if attached After the above results are determined, perform the following error-handling procedures: 1) If the Color Extended Community is invalid or is not attached (regardless of whether the Prefix-SID attribute is present or missing), the SR Policy steering actions defined in this document MUST NOT apply. 2) If both the Color Extended Community and the Prefix-SID attribute are missing or invalid: If a valid redirect-to-IP Extended Community is attached, the receiver MUST process the traffic steering solely according to (i.e., standard Redirect-to-IP). If no valid redirect-to-IP Extended Community is present, matching traffic MUST fall back to standard IP/SRv6 native shortest-path forwarding. If other valid non-steering actions (e.g., traffic-rate/drop, rate-limiting, or DSCP remarking) are attached to the FlowSpec route, the receiver MUST continue to process and apply those actions. 3) If the redirect-to-IP Extended Community is invalid or is not attached, and there are other actions attached, the filter is further processed with those actions. 4) If the redirect-to-IP Extended Community, the Color Extended Community, and the Prefix-SID attribute are attached and valid, the traffic flows are per-destination steered into the corresponding SR-Policy by the procedures with the Service SID(defined in the Prefix-SID attribute), as illustrated in Sections 3 and 5. The Headend speaker should load-share the traffic flows across all the corresponding SR-Policies with the redirect-to-IP addresses as their Endpoints if there are multiple valid redirect-to-IP Extended Communities. If the HeadEnd speaker is incapable of doing so, it SHOULD deterministically select one redirect-to-IP address as the Endpoint. 5) If the redirect-to-IP Extended Community and the Color Extended Community are attached and valid, but the Prefix-SID attribute is invalid or is not attached, the traffic flows are per-destination steered into the corresponding SR-Policy by the procedures without the Service SID, as illustrated in Sections 3 and 5. The Headend speaker should load-share the traffic flows across all the corresponding SR-Policies with the redirect-to-IP addresses as their Endpoints if there are multiple valid redirect-to-IP Extended Communities. If the HeadEnd speaker is incapable of doing so, it SHOULD deterministically select one redirect-to-IP address as the Endpoint.

In the following scenarios shown in , BGP FlowSpec Controller signals the filter rules, the Flow-spec Redirect to IPv4/IPv6 action, and the policy color to the SR-MPLS HeadEnd device.

) |TailEnd| +-------+ (SR List) +-------+ '--( )--' ( ) '-----' ]]> When the SR-MPLS HeadEnd device (as a FlowSpec client) receives such instructions from BGP FS Controller, it will steer the traffic flows matching the criteria in the FlowSpec route into the SR-MPLS Policy matching the tuple (Endpoint: TailEnd's Address, Color: C0). And the packets of such traffic flows will be encapsulated with an MPLS stack using the SR List <S1, S2, S3> in the HeadEnd device, then send the packets to the TailEnd device along the path indicated by the SR list.

In the following scenario shown in , BGP FlowSpec Controller signals the filter rules, the redirect to IPv6 Nexthop action, the policy color and the function information (SRv6 SID: Service_id_x) to the HeadEnd device.

) |TailEnd| +-------+ (SR List) +-------+ '--( )--' Service SID: Service_id_x ( ) (e.g.: End.DT4 or End.DT6 or others) '-----' ]]> When the HeadEnd device (as a FlowSpec client) receives such instructions from BGP FS Controller, it will steer the traffic flows matching the criteria in the FlowSpec route into the SRv6 Policy matching the tuple (Endpoint: TailEnd's Address, Color: C1). And the packets of such traffic flows will be encapsulated with an SRH (Segment Routing Header) using the SR List <S1, S2, S3, Service_id_x>. When the packets reach to the TailEnd device, they will be further processed per the function denoted by the Service_id_x. When the HeadEnd device determines (with the help of SRv6 SID Structure) that the Service SID belongs to the same SRv6 Locator as the last SRv6 SID of the TailEnd device in the SRv6 Policy segment list, it MAY exclude that last SRv6 SID when steering the service flow. For example, the effective segment list of the SRv6 Policy associated with SID list <S1, S2, S3> would be replaced with <S1, S2, Service_id_x>. If the last SRv6 SID (for example, we use S3 here) of the TailEnd device in the SRv6 Policy segment list is USD-flavored, an SRv6 Service SID (e.g., End.DT4 or End.DT6) is not required when a BGP FlowSpec Controller sends the FlowSpec route to the HeadEnd device (as a FlowSpec client). In the following scenario shown in , BGP FlowSpec Controller signals the filter rules, the redirect to IPv6 Nexthop action and the policy color to the HeadEnd device.

) |TailEnd| +-------+ (SR List) +-------+ '--( )--' ( ) '-----' Note: S3 MUST be a USD-flavored SRv6 SID of the TailEnd ]]> When the HeadEnd device (as a FlowSpec client) receives such instructions from a BGP FS Controller, it will steer the traffic flows matching the criteria in the Flowspec route into the SRv6 Policy matching the tuple (Endpoint: TailEnd's Address, Color: C2). And the packets of such traffic flows will be encapsulated with an SRH (Segment Routing Header) using the SR List <S1, S2, S3>. When the packets reach to the TailEnd device, they will be further processed per the function denoted by the USD-flavored SRv6 SID S3. For the cases of intra-AS and inter-AS traffic steering using this method, the usages of Flowspec Color Extended Community with BGP prefix SID are the same for both scenarios. The difference lies between the local SRv6 policy configurations. For the inter-domain case, the operator can configure an inter-domain SRv6 policy/path at the Headend device. For the intra-domain case, the operator can configure an intra-domain SRv6 policy/path at the Headend device.

[Note to the RFC Editor - remove this section before publication, as well as remove the reference to . This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in . The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist. According to , "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".

This section records the status of known implementations of the protocol defined by this specification at the time of posting of this document. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist. The Traffic Steering using BGP FlowSpec with SR-MPLS / SRv6 Policy mechanism has been implemented on the following hardware devices, Network Operating System software, and SDN controllers. They have also successfully participated in the series of joint interoperability testing events hosted by China Mobile from July 2021 to October 2021. The following hardware devices and Network Operating System software had successfully passed the interoperability testing (in alphabetical order).

As of August 2022, this feature has been deployed on the IP backbone network of China Mobile. China Mobile has fully transitioned to the SRv6 architecture, with no scenarios where SRv6 and SR-MPLSv6 coexist. As a result, when utilizing Color C and IPv6 redirect addresses, traffic steering is executed exclusively via SRv6 policies. In scenarios where an operator supports the coexistence of SRv6 and SR-MPLSv6, it is recommended to adhere to to govern policy selection for traffic steering based on Color C and IPv6 endpoints.

No IANA actions are required for this document. All code points will be used as originally assigned by IANA without any changes.

The security considerations of the base BGP FlowSpec specifications and the core Segment Routing architecture apply to this document.

The following people made significant contributions to this document:

The authors would like to acknowledge the review and inputs from Jeffrey Haas, Susan Hares, Keyur Patel, Weiqiang Cheng, Kaliraj Vairavakkalai, Robin Li, Acee Lindem, Gunter Van De Velde, John Scudder, Rainbow Wu, Linda Dunbar, Gang Yan, Feng Yang, Wim Henderickx, Robert Raszuk, Ketan Talaulikar, Changwang Lin, Aijun Wang, Hao Li, Huaimo Chen, Sheng Fang, Yuanxiang Qiu, Ran Chen, Cheng Li, Zheng Zhang, Xuewei Wang, Yanrong Liang, Xuhui Cai, Haojie Wang, Lili Wang, Nan Geng, Stephane Litkowsk and Zhenqiang Li. Special thanks to Nat Kao, who suggested adding SR-MPLS use cases to this document and provided detailed feedback on the error handling section. Special thanks to Donald E. Eastlake, 3rd, who thoroughly reviewed the entire document and made many useful suggestions for improvement.