Redundancy Policy for Redundancy
ProtectionHuawei156 Beiqing Rd.Beijing100095Chinashirley.yangfan@huawei.comHuawei156 Beiqing Rd.Beijing100095Chinagengxuesong@huawei.comHuawei156 Beiqing Rd.Beijing100095Chinazhoutianran@huawei.comVerizon Inc.gyan.s.mishra@verizon.comSPRINGThis document introduces a variant of SR Policy called Redundancy
Policy, in order to instruct the replication of service packets and
assign more than one redundancy forwarding paths used for redundancy
protection.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 .Redundancy protection is a generalized
protection mechanism by replicating and transmitting copies of flow
packets on redundancy node over multiple different and disjoint paths,
and further eliminating the redundant packets at merging node. This
document introduces Redundancy Policy to support redundancy protection,
which is a variant of SR Policy . Redundancy Policy
instructs the replication of service packets and assigns more than one
equivalent forwarding paths used for redundancy protection. Redundancy
Policy applies equally to both MPLS data plane (SR-MPLS) and Segment Routing with IPv6 data plane (SRv6) .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.The other terminologies used in this document are:Redundancy Node: the start point of redundancy protection, where the
network node replicates the flow packets.Merging Node: the end point of redundancy protection, where the
network node eliminates and ordering(optionally) the flow packets.Redundancy Policy: an extended SR Policy which instructs more than
one redundancy forwarding paths to support packet redundant
transmission.Redundancy Policy is used to enable packet replication and
instantiation more than one active ordered lists of segments between
redundancy node and merging node to steer the same flow through
different paths in an SR domain.Redundancy Policy is a variant of SR Policy and also identified
through the tuple <headend, color, endpoint>. Specifically, a
redundancy policy is identified by <redundancy node, color, merging
node>. Redundancy node is specified as IPv4/IPv6 address of headend
of Redundancy Policy, which is the node to perform packet replication.
Merging node is specified as IPv4/IPv6 address of endpoint of
Redundancy Policy, which is the node to perform packet elimination.
The value of color specifies the intent of the redundancy policy is
"redundancy protection for high reliability", which indicates service
packets are replicated into multiple copies and carried on different
forwarding paths .Redundancy policy shares the basic structure and elements with SR
Policy and its information model is shown in the following:The Redundancy Policy POL1 is identified by the tuple
<redundancy node, color, merging node>, in which R1 is the
redundancy node, M1 is the merging node, and Color 1 represents the
intent of redundancy protection. Two candidate-paths CP1 and CP2
instruct the ordered segment lists from redundancy node to merging
node. In candidate path CP1, a new attribute Flag is added to indicate
the type of candidate path. When the candidate path is indicated with
the flag of redundancy, the attribute Weight is not applicable to the
SID-Lists and all SID Lists of the candidate path are used for
redundancy forwarding. Regarding the other attributes of candidate
path such as originator, preference, priority, segment-list etc, the
definitions apply the same as .Flag is an optional attribute of a candidate path, which is used to
indicate the type of a candidate path is for redundancy forwarding.
When the candidate path with flag of redundancy is selected as the
active candidate path, this SR Policy is identified as the Redundancy
Policy. Flag of a candidate path is an 8-bit bitmap. The table below
specifies the current definition of Flag:When the SR policy is identified as a redundancy policy, network
node uses rules to compute and select the valid active ordered
segment-lists for redundancy forwarding. The specific rules are:The candidate paths are selected to determine the best path of
an SR policy. Preference, Protocol-Origin, and other tie-breaking
rules defined in section 2.9 of are evaluated
until only one valid best path is selected.In a redundancy policy, the candidate path with a flag of
redundancy is always selected as the best path in the first
place.When the selected active candidate path is with a flag of
redundancy, all the segment-lists of the candidate path are used
as the active segment-lists for redundancy forwarding, where each
active segment-list carries an entire copy of service packets.Weight is not applicable for the segment-lists in a candidate
path with a flag of redundancy. Redundancy policy has no purpose
of weighted load-balancing.The candidate path without a flag of redundancy in the same SR
policy with the candidate paths with a flag, is considered as the
backup path, which allowing provisioning of multiple path
options.Take the information model in section 3.2 as an example,
preference value 200 of CP1 is higher than preference value 100 of
CP2, thus CP1 is selected as the active candidate path. Because CP1 is
with the flag of redundancy, both Segment-List1 and Segment-List2 are
selected as the active Segment-Lists for redundancy forwarding. After
service packets are replicated, each segment-list forwards each
replicas of service packets. When CP1 becomes invalid and fallbacks to
CP2, CP2 provides the backup path to the redundancy forwarding.Redundancy policy can be optionally associated with a Binding
Segment. Redundancy SID defined in can be the Binding
SID of redundancy policy. In other words, Redundancy SID triggers the
instantiation of redundancy policy in the forwarding plane on
redundancy node.A packet is steered into a Redundancy Policy at a redundancy node
in following ways:Incoming packets have an active SID matching the Redundancy SID
at the redundancy node.Per-destination Steering: incoming packets match a BGP/Service
route which recurses on a Redundancy Policy.Per-flow Steering: incoming packets match or recurse on a
forwarding array of where some of the entries are Redundancy
Policy.Policy-based Steering: incoming packets match a routing policy
which redirects them on a Redundancy Policy.Similar to SR Policy, Redundancy Policy requires the control plane
protocol extensions to distribute candidate paths and other
information. New sub-TLVs are expected to be defined to encode new
information of Redundancy Policy Candidate Paths in BGP and PCEP .TBDTBD