Segment Routing ArchitectureCisco Systems, Inc.BrusselsBEcfilsfil@cisco.comCisco Systems, Inc.Italystefano@previdi.netCisco Systems, Incginsberg@cisco.comOrangeFRbruno.decraene@orange.comOrangeFRstephane.litkowski@orange.comGoogle, Inc.1600 Amphitheatre ParkwayMountain View, CA94043USrobjs@google.comNetwork Working GroupSegment Routing (SR) leverages the source routing paradigm. A node
steers a packet through an ordered list of instructions, called
segments. A segment can represent any instruction, topological or
service-based. A segment can have a semantic local to an SR node or
global within an SR domain. SR allows to enforce a flow through any
topological path while maintaining per-flow state only at the ingress
nodes to the SR domain.Segment Routing can be directly applied to the MPLS architecture with
no change on the forwarding plane. A segment is encoded as an MPLS
label. An ordered list of segments is encoded as a stack of labels. The
segment to process is on the top of the stack. Upon completion of a
segment, the related label is popped from the stack.Segment Routing can be applied to the IPv6 architecture, with a new
type of routing header. A segment is encoded as an IPv6 address. An
ordered list of segments is encoded as an ordered list of IPv6 addresses
in the routing header. The active segment is indicated by the
Destination Address of the packet. The next active segment is indicated
by a pointer in the new routing header.The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119.Segment Routing (SR) leverages the source routing paradigm. A node
steers a packet through an SR Policy instantiated as an ordered list of
instructions called segments. A segment can represent any instruction,
topological or service-based. A segment can have a semantic local to an
SR node or global within an SR domain. SR supports per-flow explicit
routing while maintaining per-flow state only at the ingress nodes to
the SR domain.A segment is often referred to by its Segment Identifier (SID).A segment may be associated with a topological instruction. A
topological local segment may instruct a node to forward the packet via
a specific outgoing interface. A topological global segment may instruct
an SR domain to forward the packet via a specific path to a destination.
Different segments may exist for the same destination, each with
different path objectives (e.g., which metric is minimized, what
constraints are specified).A segment may be associated with a service instruction (e.g. the
packet should be processed by a container or VM associated with the
segment). A segment may be associated with a QoS treatment (e.g., shape
the packets received with this segment at x Mbps).The SR architecture supports any type of instruction associated with
a segment.The SR architecture supports any type of control-plane: distributed,
centralized or hybrid.In a distributed scenario, the segments are allocated and signaled by
IS-IS or OSPF or BGP. A node individually decides to steer packets on a
source-routed policy (e.g., pre-computed local protection ) . A node individually
computes the source-routed policy.In a centralized scenario, the segments are allocated and
instantiated by an SR controller. The SR controller decides which nodes
need to steer which packets on which source-routed policies. The SR
controller computes the source-routed policies. The SR architecture does
not restrict how the controller programs the network. Likely options are
NETCONF, PCEP and BGP. The SR architecture does not restrict the number
of SR controllers. Specifically multiple SR controllers may program the
same SR domain. The SR architecture allows these SR controllers to
discover which SID’s are instantiated at which nodes and which
sets of local (SRLB) and global labels (SRGB) are available at which
node.A hybrid scenario complements a base distributed control-plane with a
centralized controller. For example, when the destination is outside the
IGP domain, the SR controller may compute a source-routed policy on
behalf of an IGP node. The SR architecture does not restrict how the
nodes which are part of the distributed control-plane interact with the
SR controller. Likely options are PCEP and BGP.Hosts MAY be part of an SR Domain. A centralized controller can
inform hosts about policies either by pushing these policies to hosts or
responding to requests from hosts.The SR architecture can be instantiated on various dataplanes. This
document introduces two dataplane instantiations of SR: SR over MPLS
(SR-MPLS) and SR over IPv6 (SRv6).Segment Routing can be directly applied to the MPLS architecture with
no change on the forwarding plane A segment is
encoded as an MPLS label. An SR Policy is instantiated as a stack of
labels. The segment to process (the active segment) is on the top of the
stack. Upon completion of a segment, the related label is popped from
the stack.Segment Routing can be applied to the IPv6 architecture with a new
type of routing header called the SR header (SRH) . An instruction is
associated with a segment and encoded as an IPv6 address. An SRv6
segment is also called an SRv6 SID. An SR Policy is instantiated as an
ordered list of SRv6 SID’s in the routing header. The active
segment is indicated by the Destination Address(DA) of the packet. The
next active segment is indicated by the SegmentsLeft (SL) pointer in the
SRH. When an SRv6 SID is completed, the SL is decremented and the next
segment is copied to the DA. When a packet is steered on an SR policy,
the related SRH is added to the packet.In the context of an IGP-based distributed control-plane, two
topological segments are defined: the IGP adjacency segment and the IGP
prefix segment.In the context of a BGP-based distributed control-plane, two
topological segments are defined: the BGP peering segment and the BGP
prefix segment.The headend of an SR Policy binds a SID (called Binding segment or
BSID) to its policy. When the headend receives a packet with active
segment matching the BSID of a local SR Policy, the headend steers the
packet into the associated SR Policy.This document defines the IGP, BGP and Binding segments for the
SR-MPLS and SRv6 dataplanes.Note: This document defines the architecture for Segment Routing,
including definitions of basic objects and functions and a description
of the overall design. It does NOT define the means of implementing the
architecture - that is contained in numerous referencing documents, some
of which are mentioned in this document as a convenience to the
reader.SR-MPLS: the instantiation of SR on the MPLS dataplaneSRv6: the instantiation of SR on the IPv6 dataplane.Segment: an instruction a node executes on the incoming packet (e.g.,
forward packet according to shortest path to destination, or, forward
packet through a specific interface, or, deliver the packet to a given
application/service instance).SID: a segment identifier. Note that the term SID is commonly used in
place of the term Segment, though this is technically imprecise as it
overlooks any necessary translation.SR-MPLS SID: an MPLS label or an index value into an MPLS label space
explicitly associated with the segment.SRv6 SID: an IPv6 address explicitly associated with the segment.Segment Routing Domain (SR Domain): the set of nodes participating in
the source based routing model. These nodes may be connected to the same
physical infrastructure (e.g., a Service Provider's network). They may
as well be remotely connected to each other (e.g., an enterprise VPN or
an overlay). If multiple protocol instances are deployed, the SR domain
most commonly includes all of the protocol instances in a network.
However, some deployments may wish to sub-divide the network into
multiple SR domains, each of which includes one or more protocol
instances. It is expected that all nodes in an SR Domain are managed by
the same administrative entity.Active Segment: the segment that is used by the receiving router to
process the packet. In the MPLS dataplane it is the top label. In the
IPv6 dataplane it is the destination address. .PUSH: the instruction consisting of the insertion of a segment at the
top of the segment list. In SR-MPLS the top of the segment list is the
topmost (outer) label of the label stack. In SRv6, the top of the
segment list is represented by the first segment in the Segment Routing
Header as defined in .NEXT: when the active segment is completed, NEXT is the instruction
consisting of the inspection of the next segment. The next segment
becomes active. In SR-MPLS, NEXT is implemented as a POP of the top
label. In SRv6, NEXT is implemented as the copy of the next segment from
the SRH to the Destination Address of the IPv6 header.CONTINUE: the active segment is not completed and hence remains
active. In SR-MPLS, CONTINUE instruction is implemented as a SWAP of the
top label. In SRv6, this is the plain IPv6
forwarding action of a regular IPv6 packet according to its Destination
Address.SR Global Block (SRGB): the set of global segments in the SR Domain.
If a node participates in multiple SR domains, there is one SRGB for
each SR domain. In SR-MPLS, SRGB is a local property of a node and
identifies the set of local labels reserved for global segments. In
SR-MPLS, using identical SRGBs on all nodes within the SR Domain is
strongly recommended. Doing so eases operations and troubleshooting as
the same label represents the same global segment at each node. In SRv6,
the SRGB is the set of global SRv6 SIDs in the SR Domain.SR Local Block (SRLB): local property of an SR node. If a node
participates in multiple SR domains, there is one SRLB for each SR
domain. In SR-MPLS, SRLB is a set of local labels reserved for local
segments. In SRv6, SRLB is a set of local IPv6 addresses reserved for
local SRv6 SID’s. In a controller-driven network, some controllers
or applications may use the control plane to discover the available set
of local segments.Global Segment: a segment which is part of the SRGB of the domain.
The instruction associated to the segment is defined at the SR Domain
level. A topological shortest-path segment to a given destination within
an SR domain is a typical example of a global segment.Local Segment: In SR-MPLS, this is a local label outside the SRGB. It
may be part of the explicitly advertised SRLB. In SRv6, this can be any
IPv6 address i.e., the address may be part of the SRGB but used such
that it has local significance. The instruction associated to the
segment is defined at the node level.IGP Segment: the generic name for a segment attached to a piece of
information advertised by a link-state IGP, e.g. an IGP prefix or an IGP
adjacency.IGP-Prefix Segment: an IGP-Prefix Segment is an IGP Segment
representing an IGP prefix. When an IGP-Prefix Segment is global within
the SR IGP instance/topology it identifies an instruction to forward the
packet along the path computed using the routing algorithm specified in
the algorithm field, in the topology and the IGP instance where it is
advertised. Also referred to as Prefix Segment.Prefix SID: the SID of the IGP-Prefix Segment.IGP-Anycast Segment: an IGP-Anycast Segment is an IGP-Prefix Segment
which identify an anycast prefix advertised by a set of routers.Anycast-SID: the SID of the IGP-Anycast Segment.IGP-Adjacency Segment: an IGP-Adjacency Segment is an IGP Segment
attached to a unidirectional adjacency or a set of unidirectional
adjacencies. By default, an IGP-Adjacency Segment is local (unless
explicitly advertised otherwise) to the node that advertises it. Also
referred to as Adjacency Segment.Adj-SID: the SID of the IGP-Adjacency Segment.IGP-Node Segment: an IGP-Node Segment is an IGP-Prefix Segment which
identifies a specific router (e.g., a loopback). Also referred to as
Node Segment.Node-SID: the SID of the IGP-Node Segment.SR Policy: an ordered list of segments. The headend of an SR Policy
steers packets onto the SR policy. The list of segments can be specified
explicitly in SR-MPLS as a stack of labels and in SRv6 as an ordered
list of SRv6 SID’s. Alternatively, the list of segments is
computed based on a destination and a set of optimization objective and
constraints (e.g., latency, affinity, SRLG, ...). The computation can be
local or delegated to a PCE server. An SR policy can be configured by
the operator, provisioned via NETCONF or
provisioned via PCEP . An SR policy can be used
for traffic-engineering, OAM or FRR reasons.Segment List Depth: the number of segments of an SR policy. The
entity instantiating an SR Policy at a node N should be able to discover
the depth insertion capability of the node N. For example, the PCEP SR
capability advertisement described in is one means of discovering this
capability.Forwarding Information Base (FIB): the forwarding table of a nodeWithin an SR domain, an SR-capable IGP node advertises segments for
its attached prefixes and adjacencies. These segments are called IGP
segments or IGP SIDs. They play a key role in Segment Routing and
use-cases as they enable the expression of any path throughout the SR
domain. Such a path is either expressed as a single IGP segment or a
list of multiple IGP segments.Advertisement of IGP segments requires extensions in link-state IGP
protocols. These extensions are defined in An IGP-Prefix segment is an IGP segment attached to an IGP prefix.
An IGP-Prefix segment is global (unless explicitly advertised
otherwise) within the SR domain. The context for an IGP-Prefix segment
includes the prefix, topology, and algorithm. Multiple SIDs MAY be
allocated to the same prefix so long as the tuple <prefix,
topology, algorithm> is unique.Multiple instances and topologies are defined in IS-IS and OSPF in:
, , and .Segment Routing supports the use of multiple routing algorithms
i.e, different constraint based shortest path calculations can be
supported. An algorithm identifier is included as part of a
Prefix-SID advertisement. Specification of how an algorithm specific
path calculation is done is required in the document defining the
algorithm.This document defines two algorithms:"Shortest Path": this algorithm is the default behavior. The
packet is forwarded along the well known ECMP-aware SPF
algorithm employed by the IGPs. However it is explicitly allowed
for a midpoint to implement another forwarding based on local
policy. The "Shortest Path" algorithm is in fact the default and
current behavior of most of the networks where local policies
may override the SPF decision."Strict Shortest Path (Strict-SPF)": This algorithm mandates
that the packet is forwarded according to ECMP-aware SPF
algorithm and instructs any router in the path to ignore any
possible local policy overriding the SPF decision. The SID
advertised with Strict-SPF algorithm ensures that the path the
packet is going to take is the expected, and not altered, SPF
path. Note that Fast Reroute (FRR)
mechanisms are still compliant with the Strict Shortest Path. In
other words, a packet received with a Strict-SPF SID may be
rerouted through a FRR mechanism. Strict-SPF uses the same
topology used by "Shortest Path". Obviously, nodes which do not
support Strict-SPF will not install forwarding entries for this
algorithm. Restricting the topology only to those nodes which
support this algorithm will not produce the desired forwarding
paths since the desired behavior is to follow the path
calculated by "Shortest Path". Therefore, a source SR node MUST
NOT use a source-routing policy containing a strict SPF segment
if the path crosses a node not supporting the strict-SPF
algorithm.An IGP-Prefix Segment identifies the path, to the related prefix,
computed as per the associated algorithm. A packet injected anywhere
within the SR domain with an active Prefix-SID is expected to be
forwarded along a path computed using the specified algorithm. For
this to be possible, a fully connected topology of routers
supporting the specified algorithm is required.When SR is used over the MPLS dataplane SIDs are an MPLS label or
an index into an MPLS label space (either SRGB or SRLB).Where possible, it is recommended that identical SRGBs be
configured on all nodes in an SR Domain. This simplifies
troubleshooting as the same label will be associated with the same
prefix on all nodes. In addition, it simplifies support for anycast
as detailed in Section 3.3.The following behaviors are associated with SR operating over the
MPLS dataplane:the IGP signaling extension for IGP-Prefix segment includes a
flag to indicate whether directly connected neighbors of the
node on which the prefix is attached should perform the NEXT
operation or the CONTINUE operation when processing the SID.
This behavior is equivalent to Penultimate Hop Popping (NEXT) or
Ultimate Hop Popping (CONTINUE) in MPLS.A Prefix-SID is allocated in the form of an MPLS label (or an
index in the SRGB) according to a process similar to IP address
allocation. Typically, the Prefix-SID is allocated by policy by
the operator (or NMS) and the SID very rarely changes.While SR allows to attach a local segment to an IGP prefix,
it is specifically assumed that when the terms "IGP-Prefix
Segment" and "Prefix-SID" are used, the segment is global (the
SID is allocated from the SRGB or as an index into the
advertised SRGB). This is consistent with all the described
use-cases that require global segments attached to IGP
prefixes.The allocation process MUST NOT allocate the same Prefix-SID
to different IP prefixes.If a node learns a Prefix-SID having a value that falls
outside the locally configured SRGB range, then the node MUST
NOT use the Prefix-SID and SHOULD issue an error log reporting a
misconfiguration.If a node N advertises Prefix-SID SID-R for a prefix R that
is attached to N, if N specifies CONTINUE as the operation to be
performed by directly connected neighbors, N MUST maintain the
following FIB entry:A remote node M MUST maintain the following FIB entry for any
learned Prefix-SID SID-R attached to IP prefix R: As Prefix-SIDs are specific to a given algorithm, if traffic
associated with an algorithm arrives at a node which does not
support that algorithm the traffic will be dropped as there will be
no forwarding entry matching the incoming label.When SR is used over the IPv6 dataplane: A Prefix-SID is an IPv6 address.An operator MUST explicitly instantiate an SRv6 SID. IPv6
node addresses are not SRv6 SIDs by default.A node N advertising an IPv6 address R usable as a segment
identifier MUST maintain the following FIB entry:Note that forwarding to R does not require an entry in the FIBs
of all other routers for R. Forwarding can be and most often will be
achieved by a shorter mask prefix which covers R.Independent of Segment Routing support, any remote IPv6 node will
maintain a plain IPv6 FIB entry for any prefix, no matter if the
prefix represents a segment or not. This allows forwarding of
packets to the node which owns the SID even by nodes which do not
support Segment Routing.Support of multiple algorithms applies to SRv6. Since algorithm
specific SIDs are simply IPv6 addresses, algorithm specific
forwarding entries can be achieved by assigning algorithm specific
subnets to the (set of) algorithm specific SIDs which a node
allocates.Nodes which do not support a given algorithm may still have a FIB
entry covering an algorithm specific address even though an
algorithm specific path has not been calculated by that node. This
is mitigated by the fact that nodes which do not support a given
algorithm will not be included in the topology associated with that
algorithm specific SPF and so traffic using the algorithm specific
destination will normally not flow via the excluded node. If such
traffic were to arrive and be forwarded by such a node, it will
still progress towards the destination node. The nexthop will either
be a node which supports the algorithm - in which case the packet
will be forwarded along algorithm specific paths (or be dropped if
none are available) - or the nexthop will be a node which does NOT
support the algorithm - in which case the packet will continue to be
forwarded along Algorithm 0 paths towards the destination node.An IGP Node-SID MUST NOT be associated with a prefix that is owned
by more than one router within the same routing domain.An "Anycast Segment" or "Anycast SID" enforces the ECMP-aware
shortest-path forwarding towards the closest node of the anycast set.
This is useful to express macro-engineering policies or protection
mechanisms.An IGP-Anycast segment MUST NOT reference a particular node.Within an anycast group, all routers in an SR domain MUST advertise
the same prefix with the same SID value.The figure above describes a network example with two groups of
transit devices. Group A consists of devices {A1, A2, A3 and A4}.
They are all provisioned with the anycast address 192.0.2.10/32 and
the anycast SID 100.Similarly, group B consists of devices {B1, B2, B3 and B4} and
are all provisioned with the anycast address 192.0.2.1/32, anycast
SID 200. In the above network topology, each PE device has a path to
each of the groups A and B.PE1 can choose a particular transit device group when sending
traffic to PE3 or PE4. This will be done by pushing the anycast SID
of the group in the stack.Processing the anycast, and subsequent segments, requires special
care.Considering an MPLS deployment, in the above topology, if device
PE1 (or PE2) requires to send a packet to the device PE3 (or PE4) it
needs to encapsulate the packet in an MPLS payload with the
following stack of labels. Label allocated by R1 for anycast SID 100 (outer label).Label allocated by the nearest router in group A for SID 30
(for destination PE3).While the first label is easy to compute, in this case since
there are more than one topologically nearest devices (A1 and A2),
unless A1 and A2 allocated the same label value to the same prefix,
determining the second label is impossible. Devices A1 and A2 may be
devices from different hardware vendors. If both don't allocate the
same label value for SID 30, it is impossible to use the anycast
group "A" as a transit anycast group towards PE3. Hence, PE1 (or
PE2) cannot compute an appropriate label stack to steer the packet
exclusively through the group A devices. Same holds true for devices
PE3 and PE4 when trying to send a packet to PE1 or PE2.To ease the use of anycast segment, it is recommended to
configure identical SRGBs on all nodes of a particular anycast
group. Using this method, as mentioned above, computation of the
label following the anycast segment is straightforward.Using anycast segment without configuring identical SRGBs on all
nodes belonging to the same device group may lead to misrouting (in
an MPLS VPN deployment, some traffic may leak between VPNs).The adjacency is formed by the local node (i.e., the node
advertising the adjacency in the IGP) and the remote node (i.e., the
other end of the adjacency). The local node MUST be an IGP node. The
remote node may be an adjacent IGP neighbor or a non-adjacent neighbor
(e.g., a Forwarding Adjacency, ).A packet injected anywhere within the SR domain with a segment list
{SN, SNL}, where SN is the Node-SID of node N and SNL is an Adj-SID
attached by node N to its adjacency over link L, will be forwarded
along the shortest-path to N and then be switched by N, without any IP
shortest-path consideration, towards link L. If the Adj-SID identifies
a set of adjacencies, then the node N load-balances the traffic among
the various members of the set.Similarly, when using a global Adj-SID, a packet injected anywhere
within the SR domain with a segment list {SNL}, where SNL is a global
Adj-SID attached by node N to its adjacency over link L, will be
forwarded along the shortest-path to N and then be switched by N,
without any IP shortest-path consideration, towards link L. If the
Adj-SID identifies a set of adjacencies, then the node N does
load-balance the traffic among the various members of the set. The use
of global Adj-SID allows to reduce the size of the segment list when
expressing a path at the cost of additional state (i.e.: the global
Adj-SID will be inserted by all routers within the area in their
forwarding table).An "IGP Adjacency Segment" or "Adj-SID" enforces the switching of
the packet from a node towards a defined interface or set of
interfaces. This is key to theoretically prove that any path can be
expressed as a list of segments.The encodings of the Adj-SID include a set of flags supporting the
following functionalities:Eligible for Protection (e.g., using IPFRR or MPLS-FRR).
Protection allows that in the event the interface(s) associated
with the Adj-SID are down, that the packet can still be forwarded
via an alternate path. The use of protection is clearly a policy
based decision i.e., for a given policy protection may or may not
be desirable.Indication whether the Adj-SID has local or global scope.
Default scope SHOULD be Local.Indication whether the Adj-SID is persistent across control
plane restarts. Persistence is a key attribute in ensuring that an
SR Policy does not temporarily result in misforwarding due to
reassignment of an Adj-SID.A weight (as described below) is also associated with the Adj-SID
advertisement.A node SHOULD allocate one Adj-SID for each of its adjacencies.A node MAY allocate multiple Adj-SIDs for the same adjacency. An
example is to support an Adj-SID which is eligible for protection and
an Adj-SID which is NOT eligible for protection.A node MAY associate the same Adj-SID to multiple adjacencies.In order to be able to advertise in the IGP all the Adj-SIDs
representing the IGP adjacencies between two nodes, parallel adjacency
suppression MUST NOT be performed by the IGP.When a node binds an Adj-SID to a local data-link L, the node MUST
install the following FIB entry:The Adj-SID implies, from the router advertising it, the forwarding
of the packet through the adjacency(ies) identified by the Adj-SID,
regardless of its IGP/SPF cost. In other words, the use of adjacency
segments overrides the routing decision made by the SPF algorithm.Adj-SIDs can be used in order to represent a set of parallel
interfaces between two adjacent routers.A node MUST install a FIB entry for any locally originated
adjacency segment (Adj-SID) of value W attached to a set of links B
with:When parallel adjacencies are used and associated to the same
Adj-SID, and in order to optimize the load balancing function, a
"weight" factor can be associated to the Adj-SID advertised with
each adjacency. The weight tells the ingress (or an
SDN/orchestration system) about the load-balancing factor over the
parallel adjacencies. As shown in , A
and B are connected through two parallel adjacenciesNode A advertises following Adj-SIDs and weights: Link-1: Adj-SID 1000, weight: 1Link-2: Adj-SID 1000, weight: 2Node S receives the advertisements of the parallel
adjacencies and understands that by using Adj-SID 1000 node A will
load-balance the traffic across the parallel links (link-1 and
link-2) according to a 1:2 ratio i.e., twice as many packets will
flow over Link-2 as compared to Link-1.In LAN subnetworks, link-state protocols define the concept of
Designated Router (DR, in OSPF) or Designated Intermediate System
(DIS, in IS-IS) that conduct flooding in broadcast subnetworks and
that describe the LAN topology in a special routing update (OSPF
Type2 LSA or IS-IS Pseudonode LSP).The difficulty with LANs is that each router only advertises its
connectivity to the DR/DIS and not to each of the individual nodes
in the LAN. Therefore, additional protocol mechanisms (IS-IS and
OSPF) are necessary in order for each router in the LAN to advertise
an Adj-SID associated to each neighbor in the LAN.In the following example diagram it is assumed that the all areas
are part of a single SR Domain.The example here below assumes the IPv6 control plane with the MPLS
dataplane.In area 2, node Z allocates Node-SID 150 to his local IPv6 prefix
2001:DB8::2:1/128.Area Border Routers (ABR) G and J will propagate the prefix and its
SIDs into the backbone area by creating a new instance of the prefix
according to normal inter-area/level IGP propagation rules.Nodes C and I will apply the same behavior when leaking prefixes
from the backbone area down to area 1. Therefore, node S will see
prefix 2001:DB8::2:1/128 with Prefix-SID 150 and advertised by nodes C
and I.It therefore results that a Prefix-SID remains attached to its
related IGP Prefix through the inter-area process, which is the
expected behavior in a single SR Domain.When node S sends traffic to 2001:DB8::2:1/128, it pushes
Node-SID(150) as active segment and forward it to A.When packet arrives at ABR I (or C), the ABR forwards the packet
according to the active segment (Node-SID(150)). Forwarding continues
across area borders, using the same Node-SID(150), until the packet
reaches its destination.BGP segments may be allocated and distributed by BGP.A BGP-Prefix segment is a BGP segment attached to a BGP prefix.A BGP-Prefix segment is global (unless explicitly advertised
otherwise) within the SR domain.The BGP Prefix SID is the BGP equivalent to the IGP Prefix
Segment.A likely use-case for the BGP Prefix Segment is an IGP-free
hyper-scale spine-leaf topology where connectivity is learned solely
via BGP In the context of BGP Egress Peer Engineering (EPE), as described
in , an
EPE enabled Egress PE node MAY advertise segments corresponding to its
attached peers. These segments are called BGP peering segments or BGP
peering SIDs. They enable the expression of source-routed inter-domain
paths.An ingress border router of an AS may compose a list of segments to
steer a flow along a selected path within the AS, towards a selected
egress border router C of the AS and through a specific peer. At
minimum, a BGP peering Engineering policy applied at an ingress PE
involves two segments: the Node SID of the chosen egress PE and then
the BGP peering segment for the chosen egress PE peer or peering
interface.Three types of BGP peering segments/SIDs are defined: PeerNode SID,
PeerAdj SID and PeerSet SID. PeerNode SID: a BGP PeerNode segment/SID is a local segment. At
the BGP node advertising it, its semantics is: SR header operation: NEXT.Next-Hop: the connected peering node to which the segment
is related.PeerAdj SID: a BGP PeerAdj segment/SID is a local segment. At
the BGP node advertising it, the semantic is: SR header operation: NEXT.Next-Hop: the peer connected through the interface to which
the segment is related.PeerSet SID. a BGP PeerSet segment/SID is a local segment. At
the BGP node advertising it, the semantic is: SR header operation: NEXT.Next-Hop: load-balance across any connected interface to
any peer in the related group.A peer set could be all the connected peers from the same
AS or a subset of these. A group could also span across AS. The
group definition is a policy set by the operator.The BGP extensions necessary in order to signal these BGP peering
segments are defined in In order to provide greater scalability, network opacity, and service
independence, SR utilizes a Binding SID (BSID). The BSID is bound to an
SR policy, instantiation of which may involve a list of SIDs. Any
packets received with active segment = BSID are steered onto the bound
SR Policy.A BSID may either be a local or a global SID. If local, a BSID SHOULD
be allocated from the SRLB. If global, a BSID MUST be allocated from the
SRGB.Use of a BSID allows the instantiation of the policy (the SID list)
to be stored only on the node(s) which need to impose the policy.
Direction of traffic to a node supporting the policy then only requires
imposition of the BSID. If the policy changes, this also means that only
the nodes imposing the policy need to be updated. Users of the policy
are not impacted.One use case for a Binding Segment is to provide support for an IGP
node to advertise its ability to process traffic originally destined
to another IGP node, called the Mirrored node and identified by an IP
address or a Node-SID, provided that a "Mirroring Context" segment be
inserted in the segment list prior to any service segment local to the
mirrored node.When a given node B wants to provide egress node A protection, it
advertises a segment identifying node's A context. Such segment is
called "Mirror Context Segment" and identified by the Mirror SID.The Mirror SID is advertised using the binding segment defined in
SR IGP protocol extensions .In the event of a failure, a point of local repair (PLR) diverting
traffic from A to B does a PUSH of the Mirror SID on the protected
traffic. B, when receiving the traffic with the Mirror SID as the
active segment, uses that segment and processes underlying segments in
the context of A.Segment Routing is defined for unicast. The application of the
source-route concept to Multicast is not in the scope of this
document.This document does not require any action from IANA.Segment Routing is applicable to both MPLS and IPv6 data planes.Segment Routing adds some meta-data (instructions) to the packet,
with the list of forwarding path elements (e.g., nodes, links, services,
etc.) that the packet must traverse. It has to be noted that the
complete source routed path may be represented by a single segment. This
is the case of the Binding SID.SR by default operates within a trusted domain. Traffic MUST be
filtered at the domain boundaries.The use of best practices to reduce the risk of tampering within the
trusted domain is important. Such practices are discussed in and are applicable to both SR-MPLS and SRv6.When applied to the MPLS data plane, Segment Routing does not
introduce any new behavior or any change in the way MPLS data plane
works. Therefore, from a security standpoint, this document does not
define any additional mechanism in the MPLS data plane.SR allows the expression of a source routed path using a single
segment (the Binding SID). Compared to RSVP-TE which also provides
explicit routing capability, there are no fundamental differences in
term of information provided. Both RSVP-TE and Segment Routing may
express a source routed path using a single segment.When a path is expressed using a single label, the syntax of the
meta-data is equivalent between RSVP-TE and
SR.When a source routed path is expressed with a list of segments
additional meta-data is added to the packet consisting of the source
routed path the packet must follow expressed as a segment list.When a path is expressed using a label stack, if one has access to
the meaning (i.e.: the Forwarding Equivalence Class) of the labels,
one has the knowledge of the explicit path. For the MPLS data plane,
as no data plane modification is required, there is no fundamental
change of capability. Yet, the occurrence of label stacking will
increase.SR domain boundary routers MUST filter any external traffic
destined to a label associated with a segment within the trusted
domain. This includes labels within the SRGB of the trusted domain,
labels within the SRLB of the specific boundary router, and labels
outside either of these blocks. External traffic is any traffic
received from an interface connected to a node outside the domain of
trust.From a network protection standpoint, there is an assumed trust
model such that any node imposing a label stack on a packet is assumed
to be allowed to do so. This is a significant change compared to plain
IP offering shortest path routing but not fundamentally different
compared to existing techniques providing explicit routing capability
such as RSVP-TE. By default, the explicit routing information MUST NOT
be leaked through the boundaries of the administered domain. Segment
Routing extensions that have been defined in various protocols,
leverage the security mechanisms of these protocols such as
encryption, authentication, filtering, etc.In the general case, a segment routing capable router accepts and
install labels only if these labels have been previously advertised by
a trusted source. The received information is validated using existing
control plane protocols providing authentication and security
mechanisms. Segment Routing does not define any additional security
mechanism in existing control plane protocols.Segment Routing does not introduce signaling between the source and
the mid points of a source routed path. With SR, the source routed
path is computed using SIDs previously advertised in the IP control
plane. Therefore, in addition to filtering and controlled
advertisement of SIDs at the boundaries of the SR domain, filtering in
the data plane is also required. Filtering MUST be performed on the
forwarding plane at the boundaries of the SR domain and may require
looking at multiple labels/instruction.For the MPLS data plane, there are no new requirements as the
existing MPLS architecture already allows such source routing by
stacking multiple labels. And for security protection, and already call for the
filtering of MPLS packets on trust boundaries.When applied to the IPv6 data plane, Segment Routing does introduce
the Segment Routing Header (SRH, ) which is a type of
Routing Extension header as defined in .The SRH adds some meta-data to the IPv6 packet, with the list of
forwarding path elements (e.g., nodes, links, services, etc.) that the
packet must traverse and that are represented by IPv6 addresses. A
complete source routed path may be encoded in the packet using a
single segment (single IPv6 address).SR domain boundary routers MUST filter any external traffic
destined to an address within the SRGB of the trusted domain or the
SRLB of the specific boundary router. External traffic is any traffic
received from an interface connected to a node outside the domain of
trust.From a network protection standpoint, there is an assumed trust
model such that any node adding an SRH to the packet is assumed to be
allowed to do so. Therefore, by default, the explicit routing
information MUST NOT be leaked through the boundaries of the
administered domain. Segment Routing extensions that have been defined
in various protocols, leverage the security mechanisms of these
protocols such as encryption, authentication, filtering, etc.In the general case, an SR IPv6 router accepts and install segments
identifiers (in the form of IPv6 addresses), only if these SIDs are
advertised by a trusted source. The received information is validated
using existing control plane protocols providing authentication and
security mechanisms. Segment Routing does not define any additional
security mechanism in existing control plane protocols.Problems which may arise when the above behaviors are not
implemented or when the assumed trust model is violated (e.g., through
a security breach) include:Malicious loopingEvasion of access controlsHiding the source of DOS attacksSecurity concerns with source routing at the IPv6 data plane are
more completely discussed in . The new
IPv6-based segment routing header is defined in . This document also
discusses the above security concerns.SR does not introduce new requirements for congestion control. By
default, traffic delivery is assumed to be best effort. Congestion
control may be implemented at endpoints. Where SR policies are in use
bandwidth allocation may be managed by monitoring incoming traffic
associated with the binding SID identifying the SR policy. Other
solutions such as may be applicable.In SR enabled networks, the path the packet takes is encoded in the
header. As the path is not signaled through a protocol, OAM mechanisms
are necessary in order for the network operator to validate the
effectiveness of a path as well as to check and monitor its liveness and
performance. However, it has to be noted that SR allows to reduce
substantially the number of states in transit nodes and hence the number
of elements that a transit node has to manage is smaller.SR OAM use cases for the MPLS data plane are defined in . SR OAM procedures for the MPLS
data plane are defined in .SR routers receive advertisements of SIDs (index, label or IPv6
address) from the different routing protocols being extended for SR.
Each of these protocols have monitoring and troubleshooting mechanisms
to provide operation and management functions for IP addresses that must
be extended in order to include troubleshooting and monitoring functions
of the SID.SR architecture introduces the usage of global segments. Each global
segment MUST be bound to a unique index or address within an SR domain.
The management of the allocation of such index or address by the
operator is critical for the network behavior to avoid situations like
mis-routing. In addition to the allocation policy/tooling that the
operator will have in place, an implementation SHOULD protect the
network in case of conflict detection by providing a deterministic
resolution approach.When a path is expressed using a label stack, the occurrence of label
stacking will increase. A node may want to signal in the control plane
its ability in terms of size of the label stack it can support.A YANG data model for segment routing
configuration and operations has been defined in .When Segment Routing is applied to the IPv6 data plane, segments are
identified through IPv6 addresses. The allocation, management and
troubleshooting of segment identifiers is no different than the existing
mechanisms applied to the allocation and management of IPv6
addresses.The DA of the packet gives the active segment address. The segment
list in the SRH gives the entire path of the packet. The validation of
the source routed path is done through inspection of DA and SRH present
in the packet header matched to the equivalent routing table
entries.In the context of SR over the IPv6 data plane, the source routed path
is encoded in the SRH as described in . The SR IPv6 source
routed path is instantiated into the SRH as a list of IPv6 address where
the active segment is in the Destination Address (DA) field of the IPv6
packet header. Typically, by inspecting in any node the packet header,
it is possible to derive the source routed path it belongs to. Similar
to the context of SR over MPLS data plane, an implementation may
originate path control and monitoring packets where the source routed
path is inserted in the SRH and where each segment of the path inserts
in the packet the relevant data in order to measure the end to end path
and performance.The following people have substantially contributed to the definition
of the Segment Routing architecture and to the editing of this
document:We would like to thank Dave Ward, Peter Psenak, Dan Frost, Stewart
Bryant, Pierre Francois, Thomas Telkamp, Ruediger Geib, Hannes Gredler,
Pushpasis Sarkar, Eric Rosen, Chris Bowers and Alvaro Retana for their
comments and review of this document.