SRv6 for Redundancy Protection

Redundancy Protection is a generalized protection mechanism to achieve the high reliability of service transmission in Segment Routing network. Specifically, packets of flows are replicated at a network node into two or more copies, which are transported via different and disjoint paths in parallel. When copies of packets are received and merged at one network node, the redundant packets are determined and further eliminated to guarantee only one copy of packets is transmitted. The mechanism inherits the "Live-Live" methodology, targeting to enhance the functionalities of Segment Routing over IPv6 . Inspired by DetNet Packet Replication and Packet Elimination Functions, two new Segments are introduced to provide the replication and elimination functions on specific network nodes by leveraging SRv6 Segment programming capabilities. As it is unnecessary to perform switchover of recieving packets between different paths, redundancy protection can facilitate to achieve zero packet loss target when failure on either path happens. Redundancy protection provides ultra reliable protection to many services, for example Cloud VR/Game, IPTV service and other type of video services, high value private line service etc. In this document, redundancy protection is applied to point-to-point service. The mechanism for point-to-multipoint service stays out of the scope of this document. Segment Routing (SR) leverages the source routing paradigm. An ingress node steers a packet through an ordered list of instructions, called "segments". A segment can be associated to an arbitrary processing of the packet in the node identified by the segment. This document extends the Segment Routing capabilities to support the redundancy protection in an SRv6 environment, including the definitions of two new Segments, meta data encapsulation, and a variation of Segment Routing Policy.

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 when, and only when, they appear in all capitals, as shown here.

SR: Segment Routing URLLC: Ultra-Reliable Low-Latency Communication VR: Virtual Reality Red node: Redundancy node Mer node: Merging node FID: Flow IDentification SN: Sequence Number

| | | | +-----+ | | +-----+ R3 +-----+ | | | +-----+ | | +-----+ +--+--+ +--+--+ +-----+ -------+ R1 +--------+ Red | | Mer +-------+ R2 +------- +-----+ +--+--+ +--+--+ +-----+ | +-----+ | +------+ R4 +-----+ +-----+ ]]> Figure 1: Example Scenario of Redundancy Protection in SRv6 Domain This figure shows an example of redundancy protection used in SRv6 domain. R1, R2, R3, R4, Red and Mer are SR-capable nodes. When a flow is sent into SRv6 domain, the process is: 1) R1 receives the traffic flow and encapsulates packets with a list of segments destined to R2, which is instantiated as an ordered list of SRv6 SIDs. 2) When the packet flow arrives at Red node, known as Redundancy node, each packet is replicated into two or more copies. Each copy of the packet is encapsulated with a new segment list, which represents different disjoint forwarding paths. 3) Meta data information such as flow identification (FID) and sequence number (SN) is used to facilitate the packet elimination on Merging node (Mer). Flow identification identifies the specific flow, and sequence number distinguishes the packet sequence of a flow. Meta data is either carried in the packet before it arrives at redundancy node, or added to each of the replicas at redundancy node. 4) The multiple replicas go through different paths until the Mer node, known as Merging node. The first received copy of each flow packet is transmitted from Merging node to R2, and the redundant packets are eliminated. 5) When there is any failures or packet loss in one path, the service transmission continues through the other path non-disruptively. 6) Sometimes, out-of-order packets may occur since service packets are recovered from different forwarding paths. In this case, the merging node or other network nodes behind merging node is desired to include a reordering function, which is implementation specific and out of the scope of this document. In this example, service protection is supported by utilizing two packet flows transmitted over two forwarding paths. It is noted that there is no limitation of the number of replicas. For a unidirectional flow, Red node supports replication function, and Mer node supports elimination function. Reordering function MAY be required in combination of elimination function on merging node. To minimize the jitter caused by random packet loss, the disjoint paths are recommended to have similar path forwarding delay.

To achieve the packet replication and elimination functions, Redundancy Segment and Merging Segment, as well as the related SRv6 Endpoint Behavior are introduced.

Redundancy Segment is the identifier of packets which need to be replicated on redundancy node. It is a variation of Binding SID (BSID) to associate with a Redundancy Policy, instantiation of which provides segment lists of different disjoint paths. Similar to the relationship between BSID and SR Policy , the use of Redundancy Segment would trigger the Redundancy Policy instantiation on redundancy node. Redundancy Segment is associated with service instructions, indicating the following operations: Steers the packet into the corresponding redundancy policy Encapsulates flow identification and sequence number in packets if the two information is not carried in packets Packet replication and segment encapsulation based on the information of redundancy policy, e.g., the number of replication copies, an ordered list of segments with a topological instruction In the case of SRv6, a new behavior End.R for Redundancy Segment is defined. An instance of a redundancy SID is associated with a redundancy policy B and a source address A. In the following description, End.R behavior is specified in the encapsulation mode. The End.R behavior in the insertion mode is for further study. When an SRv6-capable node (N) receives an IPv6 packet whose destination address matches a local IPv6 address instantiated as an SRv6 SID (S), and S is a Redundancy SID, N does:

0) { S03. Decrement IPv6 Hop Limit by 1 S04. Decrement Segments Left by 1 S05. Update IPv6 DA with Segment List[Segments Left] S06. Add flow identification and sequence number if indicated* S07. Duplicate the packets (as number of active SID lists in B) S08. Push the new IPv6 headers to each replica. The IPv6 header contains an SRH with the SID list in B S09. Set the outer IPv6 SA to A S10. Set the outer IPv6 DA to the first SID of new SRH SL S11. Set the outer Payload Length, Traffic Class, Flow Label, Hop Limit and Next-Header fields S12. Submit the packet to the egress IPv6 FIB lookup for transmission to the new destination S13. } S14. } * Adding flow identification and sequence number is an optional behavior for Redundancy Segment. The instruction execution is determined and explicitly indicated by SR policy or Segment itself. ]]>

Merging Segment is associated with service instructions, indicates the following operations: Packet merging and elimination: forward the first received packets and eliminate the redundant packets In order to eliminate the redundant packet of a flow, merging node utilizes sequence number to evaluate the redundant status of a packet. Note that implementation specific mechanism could be applied to control the amount of state monitored on sequence number, so that system memory usage can be limited at a reasonable level. As merging node needs to maintain the state of flows, a centralized controller should have a knowledge of merging nodes capability, and never provision the redundancy policy to redundancy node when the computation result goes beyond the flow recovery capability of merging node. The capability advertisement of merging node will be specified separately elsewhere, which is not within the scope of this document. In the case of SRv6, a new behavior End.M for Merging Segment is defined. When an SRv6-capable node (N) receives an IPv6 packet whose destination address matches a local IPv6 address instantiated as an SRv6 SID (S), and S is a Merging SID, N does:

or ==0) { S03. Acquire the sequence number of received packet and look it up in table S04. If (this sequence number does not exist in the table) { S05. Store this sequence number in table S06. Remove the outer IPv6+SRH header S07. Decrement IPv6 Hop Limit by 1 in inner SRH S08. Decrement Segments Left by 1 in inner SRH S09. Update IPv6 DA with Segment List[Segments Left] in inner SRH S10. Submit the packet to the egress IPv6 FIB lookup and transmit S11. } S12. ELSE { S13. Drop the packet S14. } S15. } S16. } ]]>

To support the redundancy protection function, flow identification and sequence number are added in the packet and further used at merging node when the merging function is executed. Flow identification identifies one specific flow of redundancy protection, and is usually allocated from centralized controller to SR ingress node or redundancy node in SR network. Note that flow identification can also be allocated and advertised by merging node. BGP, PCEP or Netconf protocols can facilitate the advertisement and distribution of flow identification among controller, redundancy node and merging node. Sequence number distinguishes the packets within a flow by specifying the order of packets. Not like the uniqueness of flow identification to one specific flow, sequence number keeps changing to each packet within a flow. It is RECOMMENDED to add the sequence number in forwarding plane as performance and scalability is required. Figure 4 suggests an encapsulation of flow identification and sequence number in Segment Routing Header (SRH) when redundancy protection is used in SRv6 network.

Figure 4 Encapsulation of Flow Identification and Sequence Number Since the flow identification is only used at merging node to identify the specific flow of redundancy protection, it is RECOMMENDED to be encapsulated in the Arguments of Merging Segment in SRH. The length of flow identification is not limited, however in practice it is suggested to be 16 bits. All the duplicates of the same packet need to be tagged for deduplication at the merging node. For this purpose, we will use a sequence number. It is RECOMMENDED to encode the seq number in the Tag field of the SRH, with a length of 16bits.

Redundancy Policy is a variation of SR Policy to conduct the replicas to multiple disjoint paths for redundancy protection. It extends SR policy to include more than one active and parallel ordered lists of segments between redundancy node and merging node, and all the ordered lists of segments are used at the same time to steer each copy of flow into different disjoint paths.

This document requires registration of End.R behavior and End.M behavior in "SRv6 Endpoint Behaviors" sub-registry of "Segment Routing Parameters" registry.

TBD

The authors would like to thank Bruno Decraene, Ron Bonica, James Guichard, Jeffrey Zhang, Balazs Varga for their valuable comments and discussions.