SRv6 for Inter-Layer Network Programming

In many network scenarios, operator owns a multi-layered network. In layer-3, the technology has converged to IP, while there can be different technologies in layer-2 and below. In such networks, the cross-layer planning and optimization is considered more efficient than independent planning and operation of the layer-3 and the underlying networks in terms of resource utilization and SLA assurance, but it is also considered more complicated. Thus a mechanism for flexible and efficient inter-layer network integration is desired. Segment Routing over IPv6 (SRv6) enables a network operator or an application to specify a packet processing program by encoding a sequence of instructions in the IPv6 packet header. Currently SRv6 does not consider about the network layers under the IP layer. However, with the capability of SRv6 network programming, it is possible to achieve seamless integration between IP (layer-3) and the underlying (layer-2 and below) networks. Following the SRv6 network programming concept, a new SRv6 behavior called End.XU is defined for sending packets through an underlay interface, which connects to a remote layer-3 node via underlying links or connections. The SRv6 End.XU behavior can be considered as a variant of the SRv6 END.X behavior as defined in . Unlike an L3 adjacency, the underlying links or connections can be unidirectional and does not require bidirectional check. Thus the underlay links or connections are invisible in the L3 topology and will not be used for IP distributed route computation (e.g. SPF). However, this may just be the expected behavior in inter-layer programming where the underlay links or connections are provisioned for traffic engineering for specific types of services. Such underlying links or connections may be realized using either Metro Transport Network (MTN) paths , or ODUk or DWDM connections. The SRv6 End.XU SIDs can be used together with other types of SRv6 SIDs to build SRv6 SID lists for inter-layer network programming. This document first describes the typical use cases of inter-layer network programming, then a new SRv6 End.XU behavior for inter-layer network programming is introduced. The applicability of SRv6 End.XU behavior in typical inter-layer network programming scenarios is also illustrated.

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.

In many network scenarios, the underlay of the IP network is an optical network. The IP network and optical network are usually managed separately, the optical network works as an underlay which is normally invisible to the IP network. In some cases, the optical path resources and the IP path resources may not be one-to-one mapping, which makes the redundant optical paths not fully used by the IP layer. In some other cases, there may be optical paths between non-adjacent IP nodes thus they are not visible in the L3 topology, thus they can not be used for carrying traffic based on IP routing. However, such optical paths may be used for inter-layer traffic engineering.

The architecture of Metro Transport Network (MTN) is defined in . In an MTN based network, network nodes can support two forwarding modes: per-hop IP packet forwarding and the MTN Path (MTNP) layer cross-connect. An MTN path is a multi-hop underlay transport path which may be established between any two nodes in the MTN network, and the intermediate nodes on the MTN path will forward the traffic based on the pre-established MTN cross-connect without IP table lookup. Thus an MTN path is considered as an underlay connection between two remote MTN nodes. Although in some cases it is possible to set up a layer-3 adjacency between the two endpoints of the MTN path, it will make the provisioning of MTN path complicated. Moreover, in some cases the two endpoints may reside in different IGP areas or ASes, which makes a layer-3 adjacency between them more challenging. Last but not the least, the MTN path may be provisioned unidirectionally, which cannot pass the bidirectional connectivity check required for a layer-3 link. Since the MTN paths are usually not visible in the L3 topology, it is difficult to compute and establish an end-to-end inter-layer path which consists of both the layer-3 network segments and the MTN paths.

This section defines a new SRv6 behavior for the underlay cross-connect. The "Endpoint with Underlay cross-connect" behavior ("End.XU" for short) is a variant of the End.X behavior defined in . Its main use is for SRv6 based inter-layer network programming and traffic engineering. Any SID instance of this behavior is associated with an underlay interface, which connects to one or more underlay links or connections. When node N receives a packet destined to S and S is a local End.XU SID, N does the following:

max_LE) or (Segments Left > Last Entry+1)) { S10. Send an ICMP Parameter Problem to the Source Address with Code 0 (Erroneous header field encountered) and Pointer set to the Segments Left field, interrupt packet processing, and discard the packet. S11. } S12. Decrement IPv6 Hop Limit by 1 S13. Decrement Segments Left by 1 S14. Update IPv6 DA with Segment List[Segments Left] S15. Send the packet through one of the underlay links associated with the underlay interface identified by S. S16. } ]]> Note that the underlay interface and the associated links in step 15 SHOULD be established before the associated End.XU SID is announced into the network. When forwarding packets through the underlay interface towards the remote endpoint node, the information required for layer-2 encapsulation may be provisioned via mechanisms such as static Neighbor Discovery (ND) Cache. The details are out of the scope of this document. End.XU SIDs MAY be announced using IGP or BGP-LS in a similar way to the announcement of the End.X SIDs, while the information about the underlay connections and the associated End.XU SIDs need to be distinguished from the layer-3 links and the End.X SIDs. The detailed protocol extensions will be described in a separate document. With the collected information of End.XU SIDs, the network controller or headend nodes could use the End.XU SIDs together with other types of SRv6 SIDs to build SRv6 SID lists for inter-layer TE paths.

Assuming that an operator owns both the IP and optical network, and the operator needs to deploy E2E service across IP and optical network, with traditional approaches the planning and service provisioning would be complex and time consuming due of the manual synergy needed between the operator's IP team and optical team. With the introduction of SRv6 and the End.XU behavior, one simplified approach for IP and optical integration is to build a SRv6 SID list that integrates the path in both the IP layer and the optical layer. As the optical layer is not packet based, source routing mechanism can not be directly used in the optical network. However, the abstracted optical paths (e.g., with ODUk or DWDM) could be exposed to the control system of the IP network using the SRv6 End.XU SIDs, and some of the attributes of the optical paths may also be provided. Based on this information, IP-optical inter-layer paths can be computed and programmed to meet some specific service requirements, such as low latency.

In Figure 1, P1 to P8 are IP nodes, and O1 to O6 are optical nodes. Assume the operator needs to deploy a low latency path between P7 and P8. With normal segment routing, an IP layer path with the segment list {P7, P1, P2, P3, P8} can be used. But if an optical path from O1 to O3 exists, and the End.XU SID defined in this document is used to announce this optical path as an underlay connection with specific attributes into the IP network, the headend node or the controller in IP layer can program an inter-layer TE path along {P7, P1, End.XU (O1, O2, O3), P3, P8} which may provide lower latency. The optical path between O1 and O3 may be created in advance or as a result of the request from the IP layer. The creation should be done by the optical network controller (not shown in the figure). The details of the process are out of scope of this document, and may refer to . There is also another case of IP and Optical integration. Assume there are two optical paths between P1 and P2. One is {P1, O1, O2, P2} , and the other is {P1, O1, O4, O5, O2, P2}. Two separate End.XU SIDs can be allocated for these two underlay connections respectively. One is End.XU P1::C2 for the underlay path {P1, O1, O2, P2}, and the other is End.XU P1::C45 for the path {P1, O1, O4, O5, O2, P2}. The headend P7 or the IP network controller will be informed about these two SRv6 End.XU SIDs and the associated path attributes, so that the headend or the controller can program different end-to-end inter-layer paths using SRv6 SID lists with different End.XU SIDs for services with different SLA requirements.

Assuming that an operator owns both an MTN network domain and an IP network domain. In the MTN network, each MTN node has both the layer-3 functionality and the MTN Path layer functionality. In layer-3, all the MTN nodes are in a layer-3 network topology, which connects to the IP network domain. In the MTN Path Layer, a set of MTN paths are provisioned between the selected pairs of MTN nodes for traffic engineering. In the MTN network, different types of services may be carried using either a layer-3 path, an end-to-end MTN path, or an inter-layer path comprising of both the layer-3 links and the MTN paths as segments. In addition, For some type of services, end-to-end paths across the IP domain and the MTN domain are needed, which is comprised of both the layer-3 paths and the MTN path as different segments.

Figure 2 gives an example of a network with a MTN domain and an IP domain. M1 to M7 are MTN nodes, and P1 to P4 are IP nodes. The same set of MTN nodes builds two separate network layers. The topology in the IP layer shows the layer-3 connectivity between the MTN nodes and the connectivity with the IP network domain, while the topology in the MTN Path layer shows the MTN paths between the selected pair of MTN nodes. An end-to-end path from M7 to P5 can be established in layer-3 using an SRv6 SID list representing the layer-3 path {M7, M1, M2, M3, P1, P2, P5}. While for services which require low latency, an end-to-end path consisting of both the layer-3 segments and MTN paths could be established using an SRv6 SID list representing the inter-layer path {M7, M1::C3, P1, P2, P5}, where the End.XU SID M1::C3 represents the MTN path M1'-M3'. This shows that it is convenient to use integrated SRv6 SID lists to program inter-layer TE paths both within the MTN domain, and across the IP and MTN domain using the combination of L3 SRv6 SIDs and the End.XU SIDs.

TBD

This document defines a new SRv6 Endpoint behavior called END.XU. IANA has allocated the following code points for different flavors of End.XU from the "SRv6 Endpoint Behaviors" sub-registry in the "Segment-routing with IPv6 data plane (SRv6) Parameters" registry:

The authors would like to thank Xiaodong Chang, Yongjian Hu, Alexander Vainshtein, Ketan Talaulikar and Zhibo Hu for their review and comments.