Internet DRAFT - draft-ietf-mpls-tp-shared-ring-protection
draft-ietf-mpls-tp-shared-ring-protection
Network Working Group W. Cheng
Internet-Draft L. Wang
Intended status: Standards Track H. Li
Expires: December 14, 2017 China Mobile
H. Helvoort
Hai Gaoming BV
J. Dong
Huawei Technologies
June 12, 2017
Shared-Ring protection (MSRP) mechanism for ring topology
draft-ietf-mpls-tp-shared-ring-protection-06
Abstract
This document describes requirements, architecture and solutions for
MPLS-TP Shared Ring Protection (MSRP) in a ring topology for point-
to-point (P2P) services. The MSRP mechanism is described to meet the
ring protection requirements as described in RFC 5654. This document
defines the Ring Protection Switch (RPS) Protocol that is used to
coordinate the protection behavior of the nodes on MPLS ring.
Requirements Language
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 [RFC2119].
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on December 14, 2017.
Cheng, et al. Expires December 14, 2017 [Page 1]
Internet-Draft MSRP June 2017
Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology and Notation . . . . . . . . . . . . . . . . . . 3
3. MPLS-TP Ring Protection Criteria and Requirements . . . . . . 4
4. Shared Ring Protection Architecture . . . . . . . . . . . . . 5
4.1. Ring Tunnel . . . . . . . . . . . . . . . . . . . . . . . 5
4.1.1. Establishment of Ring Tunnel . . . . . . . . . . . . 6
4.1.2. Label Assignment and Distribution . . . . . . . . . . 8
4.1.3. Forwarding Operation . . . . . . . . . . . . . . . . 8
4.2. Failure Detection . . . . . . . . . . . . . . . . . . . . 9
4.3. Ring Protection . . . . . . . . . . . . . . . . . . . . . 10
4.3.1. Wrapping . . . . . . . . . . . . . . . . . . . . . . 11
4.3.2. Short Wrapping . . . . . . . . . . . . . . . . . . . 13
4.3.3. Steering . . . . . . . . . . . . . . . . . . . . . . 16
4.4. Interconnected Ring Protection . . . . . . . . . . . . . 19
4.4.1. Interconnected Ring Topology . . . . . . . . . . . . 19
4.4.2. Interconnected Ring Protection Mechanisms . . . . . . 21
4.4.3. Ring Tunnels in Interconnected Rings . . . . . . . . 22
4.4.4. Interconnected Ring Switching Procedure . . . . . . . 24
4.4.5. Interconnected Ring Detection Mechanism . . . . . . . 25
5. Ring Protection Coordination Protocol . . . . . . . . . . . . 26
5.1. RPS and PSC Comparison on Ring Topology . . . . . . . . . 26
5.2. RPS Protocol . . . . . . . . . . . . . . . . . . . . . . 27
5.2.1. Transmission and Acceptance of RPS Requests . . . . . 29
5.2.2. RPS PDU Format . . . . . . . . . . . . . . . . . . . 29
5.2.3. Ring Node RPS States . . . . . . . . . . . . . . . . 31
5.2.4. RPS State Transitions . . . . . . . . . . . . . . . . 33
5.3. RPS State Machine . . . . . . . . . . . . . . . . . . . . 35
5.3.1. Switch Initiation Criteria . . . . . . . . . . . . . 35
5.3.2. Initial States . . . . . . . . . . . . . . . . . . . 37
5.3.3. State transitions When Local Request is Applied . . . 38
5.3.4. State Transitions When Remote Request is Applied . . 42
Cheng, et al. Expires December 14, 2017 [Page 2]
Internet-Draft MSRP June 2017
5.3.5. State Transitions When Request Addresses to Another
Node is Received . . . . . . . . . . . . . . . . . . 45
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 47
6.1. G-ACh Channel Type . . . . . . . . . . . . . . . . . . . 48
6.2. RPS Request Codes . . . . . . . . . . . . . . . . . . . . 48
7. Operational Considerations . . . . . . . . . . . . . . . . . 48
8. Security Considerations . . . . . . . . . . . . . . . . . . . 49
9. Contributing Authors . . . . . . . . . . . . . . . . . . . . 50
10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 51
11. References . . . . . . . . . . . . . . . . . . . . . . . . . 51
11.1. Normative References . . . . . . . . . . . . . . . . . . 52
11.2. Informative References . . . . . . . . . . . . . . . . . 52
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 53
1. Introduction
As described in section 2.5.6.1 of [RFC5654], several service
providers have expressed much interest in operating MPLS-TP in ring
topologies and require a high- level survivability function in these
topologies. In operational transport network deployment, MPLS-TP
networks are often constructed using ring topologies. This calls for
an efficient and optimized ring protection mechanism to achieve
simple operation and fast, sub 50 ms, recovery performance.
This document specifies an MPLS-TP Shared-Ring Protection mechanisms
that meets the criteria for ring protection and the ring protection
requirements described in section 2.5.6.1 of [RFC5654].
The basic concept and architecture of the Shared-Ring protection
mechanism are specified in this document. This document describes
the solutions for point-to-point transport paths. While the basic
concept may also apply to point-to-multipoint transport paths, the
solution for point-to-multipoint transport paths is out of the scope
of this document.
2. Terminology and Notation
Terminology:
Ring Node: All nodes in the ring topology are Ring Nodes and they
MUST actively participate in the ring protection.
Ring tunnel: A ring tunnel provides a server layer for the LSPs
traversing the ring. The notation used for a ring tunnel is:
R<d><p><X> where <d> = c (clockwise) or a (anticlockwise), <p> = W
(working) or P (protecting), and <X> = the node name.
Cheng, et al. Expires December 14, 2017 [Page 3]
Internet-Draft MSRP June 2017
Ring map: A ring map is present in each ring node. The ring-map
contains the ring topology information, i.e. the nodes in the ring,
the adjacency of the ring nodes and the status of the links between
ring nodes (Intact or Severed). The ring map is used by every ring
node to determine the switchover behavior of the ring tunnels.
Notation:
The following syntax will be used to describe the contents of the
label stack:
1. The label stack will be enclosed in square brackets ("[]").
2. Each level in the stack will be separated by the '|' character.
It should be noted that the label stack may contain additional
layers. However, we only present the layers that are related to the
protection mechanism.
3. If the Label is assigned by Node X, the Node Name is enclosed in
parentheses ("()").
3. MPLS-TP Ring Protection Criteria and Requirements
The generic requirements for MPLS-TP protection are specified in
[RFC5654]. The requirements specific for ring protection are
specified in section 2.5.6.1 of [RFC5654]. This section describes
how the criteria for ring protection are met:
a. The number of OAM entities needed to trigger protection
Each ring node requires only one instance of the RPS protocol per
ring. The OAM of the links connected to the adjacent ring-nodes has
to be forwarded to only this instance in order to trigger protection.
For detailed information, see section 5.2.
b. The number of elements of recovery in the ring
Each ring-node requires only one instance of the RPS protocol and is
independent of the number of LSPs that are protected. For detailed
information, see section 5.2.
c. The required number of labels required for the protection paths
The RPS protocol uses ring tunnels and each tunnel has a set of
labels. The number of ring tunnel labels is related to the number of
ring-nodes and is independent of the number of protected LSPs. For
detailed information, see section 4.1.2.
Cheng, et al. Expires December 14, 2017 [Page 4]
Internet-Draft MSRP June 2017
d. The amount of control and management-plane transactions
Each ring-node requires only one instance of the RPS protocol per
ring. This means that only one maintenance operation is required per
ring-node. For detailed information, see section 5.2.
e. Minimize the signaling and routing information exchange during
protection
Information exchange during a protection switch is using the in-band
RPS and OAM messages. No control plane interactions are required.
For detailed information, see section 5.2.
4. Shared Ring Protection Architecture
4.1. Ring Tunnel
This document introduces a new logical layer of the ring for shared
ring protection in MPLS-TP networks. As shown in Figure 1, the new
logical layer consists of ring tunnels which provides a server layer
for the LSPs traversing the ring. Once a ring tunnel is established,
the forwarding and protection switching of the ring are all performed
at the ring tunnel level. A port can carry multiple ring tunnels,
and a ring tunnel can carry multiple LSPs.
Cheng, et al. Expires December 14, 2017 [Page 5]
Internet-Draft MSRP June 2017
+-------------
+-------------|
+-------------| |
===Service1===| | |
===Service2===| LSP1 | |
+-------------| |
|Ring-Tunnel1 |
+-------------| |
===Service3===| | |
===Service4===| LSP2 | |
+-------------| |
+-------------| Physical
+-------------|
+-------------| | Port
===Service5===| | |
===Service6===| LSP3 | |
+-------------| |
|Ring-Tunnel2 |
+-------------| |
===Service7===| | |
===Service8===| LSP4 | |
+-------------| |
+-------------|
+-------------
Figure 1. The logical layers of the ring
The label stack used in MPLS-TP Shared Ring Protection mechanism is
[Ring Tunnel Label|LSP Label|service Label](Payload) as illustrated
in figure 2.
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Ring tunnel Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| LSP Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Service Label |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 2. Label stack used in MPLS-TP Shared Ring Protection
4.1.1. Establishment of Ring Tunnel
The Ring tunnels are established based on the egress nodes. The
egress node is the node where traffic leaves the ring. LSPs which
have the same egress node on the ring and travels along the ring in
the same direction (clockwise or anticlockwise) share the same ring
Cheng, et al. Expires December 14, 2017 [Page 6]
Internet-Draft MSRP June 2017
tunnels. In other words, all the LSPs that traverse the ring in the
same direction and exit from the same node share the same working
ring tunnel and protection ring tunnel. For each egress node, four
ring tunnels are established:
o one clockwise working ring tunnel, which is protected by the
anticlockwise protection ring tunnel
o one anticlockwise protection ring tunnel
o one anticlockwise working ring tunnel, which is protected by the
clockwise protection ring tunnel
o one clockwise protection ring tunnel
The structure of the protection tunnels is determined by the selected
protection mechanism. This will be detailed in subsequent sections.
As shown in Figure 3, LSP1, LSP2 and LSP3 enter the ring from Node E,
Node A and Node B respectively, and all leave the ring at Node D. To
protect these LSPs that traverse the ring, a clockwise working ring
tunnel (RcW_D) via E->F->A->B->C->D, and its anticlockwise protection
ring tunnel (RaP_D) via D->C->B->A->F->E->D are established, Also, an
anti-clockwise working ring tunnel (RaW_D) via C->B->A->F->E->D, and
its clockwise protection ring tunnel (RcP_D) via D->E->F->A->B->C->D
are established. For simplicity Figure 3 only shows RcW_D and RaP_D.
A similar provisioning should be applied for any other node on the
ring. In summary, for each node in Figure 3 when acting as egress
node, the ring tunnels are created as follows:
o To Node A: RcW_A, RaW_A, RcP_A, RaP_A
o To Node B: RcW_B, RaW_B, RcP_B, RaP_B
o To Node C: RcW_C, RaW_C, RcP_C, RaP_C
o To Node D: RcW_D, RaW_D, RcP_D, RaP_D
o To Node E: RcW_E, RaW_E, RcP_E, RaP_E
o To Node F: RcW_F, RaW_F, RcP_F, RaP_F
Cheng, et al. Expires December 14, 2017 [Page 7]
Internet-Draft MSRP June 2017
+---+#############+---+
| F |-------------| A | +-- LSP2
+---+*************+---+
#/* *\#
#/* *\#
#/* *\#
+---+ +---+
LSP1-+ | E | | B |+-- LSP3
+---+ +---+
#\ */#
#\ */#
#\ */#
+---+*************+---+
LSP1 +--| D |-------------| C |
LSP2 +---+#############+---+
LSP3
---- physical links
**** RcW_D
#### RaP_D
Figure 3. Ring tunnels in MSRP
Through these working and protection ring tunnels, LSPs which enter
the ring from any node can reach any egress nodes on the ring, and
are protected from failures on the ring.
4.1.2. Label Assignment and Distribution
The ring tunnel labels are downstream-assigned labels as defined in
[RFC3031]. The ring tunnel labels on each hop of the ring tunnel can
be either configured statically, provisioned by a controller, or
distributed dynamically via a control protocol. For an LSP which
traverses the ring tunnel, the ingress ring node and the egress ring
node are considered adjacent at the LSP layer, and LSP label needs to
be allocated at these two ring nodes. The control plane for label
distribution is outside the scope of this document.
4.1.3. Forwarding Operation
When an MPLS-TP transport path, i.e. an LSP, enters the ring, the
ingress node on the ring pushes the working ring tunnel label which
is used to reach the specific egress node and sends the traffic to
the next hop. The transit nodes on the working ring tunnel swap the
ring tunnel labels and forward the packets to the next hop. When the
packet arrives at the egress node, the egress node pops the ring
tunnel label and forwards the packets based on the inner LSP label
and service label. Figure 4 shows the label operation in the MPLS-TP
Cheng, et al. Expires December 14, 2017 [Page 8]
Internet-Draft MSRP June 2017
shared ring protection mechanism. Assume that LSP1 enters the ring
at Node A and exits from Node D, and the following label operations
are executed.
1. Ingress node: Packets of LSP1 arrive at Node A with a label stack
[LSP1] and are supposed to be forwarded in the clockwise
direction of the ring. The label of the clockwise working ring
tunnel RcW_D will be pushed at Node A, the label stack for the
forwarded packet at Node A is changed to [RcW_D(B)|LSP1].
2. Transit nodes: In this case, Node B and Node C forward the
packets by swapping the working ring tunnel labels. For example,
the label [RcW_D(B)|LSP1] is swapped to [RcW_D(C)|LSP1] at Node
B.
3. Egress node: When the packet arrives at Node D (i.e. the egress
node) with label stack [RcW_D(D)|LSP1], Node D pops RcW_D(D), and
subsequently deals with the inner labels of LSP1.
+---+#####[RaP_D(F)]######+---+
| F |---------------------| A | +-- LSP1
+---+*****[RcW_D(A)]******+---+
#/* *\#
[RaP_D(E)]#/*[RcW_D(F)] [RcW_D(B)]*\#[RaP_D(A)]
#/* *\#
+---+ +---+
| E | | B |
+---+ +---+
#\ */#
[RaP_D(D)]#\ [RxW_D(C)]*/#[RaP_D(B)]
#\ */#
+---+*****[RcW_D(D)]****+---+
LSP1 +-- | D |-------------------| C |
+---+#####[RaP_D(C)]####+---+
-----physical links ****** RcW_D ###### RaP_D
Figure 4. Label operation of MSRP
4.2. Failure Detection
The MPLS-TP section layer OAM is used to monitor the connectivity
between each two adjacent nodes on the ring using the mechanisms
defined in [RFC6371]. Protection switching is triggered by the
failure detected on the ring by the OAM mechanisms.
Two ports of a link form a Maintenance Entity Group (MEG), and an MEG
end point (MEP) function is installed in each ring port. CC OAM
Cheng, et al. Expires December 14, 2017 [Page 9]
Internet-Draft MSRP June 2017
packets are periodically exchanged between each pair of MEPs to
monitor the link health. Three consecutive lost CC packets MUST be
interpreted as a link failure.
A node failure is regarded as the failure of two links attached to
that node. The two nodes adjacent to the failed node detect the
failure in the links that are connected to the failed node.
4.3. Ring Protection
This section specifies the ring protection mechanisms in detail. In
general, the description uses the clockwise working ring tunnel and
the corresponding anti-clockwise protection ring tunnel as an
example, but the mechanism is applicable in the same way to the anti-
clockwise working and clockwise protection ring tunnels.
In a ring network, each working ring tunnel is associated with a
protection ring tunnel in the opposite direction, and every node MUST
obtain the ring topology either by configuration or via a topology
discovery mechanism. The ring topology and the connectivity (Intact
or Severed) between two adjacent ring nodes form the ring map. Each
ring node maintains the ring map and uses it to perform ring
protection switching.
Taking the topology in Figure 4 as an example, LSP1 enters the ring
at Node A and leaves the ring at Node D. In normal state, LSP1 is
carried by the clockwise working ring tunnel (RcW_D) through the path
A->B->C->D. The label operation is:
[LSP1](Payload) -> [RCW_D(B)|LSP1](NodeA) -> [RCW_D(C)|LSP1](NodeB)
-> [RCW_D(D)| LSP1](NodeC) -> [LSP1](Payload).
Then at node D the packet will be forwarded based on the label stack
of LSP1.
Three typical ring protection mechanisms are described in this
section: wrapping, short wrapping and steering. All nodes on the
same ring MUST use the same protection mechanism. If the RPS
protocol in any node detects RPS message with a protection switching
mode that was not provisioned in that node a failure of protocol will
be reported, and the protection mechanism will not be activated.
Wrapping ring protection: the node which detects a failure or accepts
a switch request switches the traffic impacted by the failure or the
switch request to the opposite direction (away from the failure). In
this way, the impacted traffic is switched to the protection ring
tunnel by the switching node upstream of the failure, then travels
around the ring to the switching node downstream of the failure
Cheng, et al. Expires December 14, 2017 [Page 10]
Internet-Draft MSRP June 2017
through the protection ring tunnel, where it is switched back onto
the working ring tunnel to reach the egress node.
Short wrapping ring protection provides some optimization to wrapping
protection, in which the impacted traffic is only switched once to
the protection ring tunnel by the switching node upstream to the
failure. At the egress node, the traffic leave the ring from the
protection ring tunnel. This can reduce the traffic detour of
wrapping protection.
Steering ring protection implies that the node that detects a failure
sends a request along the ring to the other node adjacent to the
failure, and all nodes in the ring process this information. For the
impacted traffic, the ingress node (which adds traffic to the ring)
performs switching of the traffic from working to the protection ring
tunnel, and the egress node will drop the traffic received from the
protection ring tunnel.
The following sections describe these protection mechanisms in
detail.
4.3.1. Wrapping
With the wrapping mechanism, the protection ring tunnel is a closed
ring identified by the egress node. As shown in Figure 4, the RaP_D
is the anticlockwise protection ring tunnel for the clockwise working
ring tunnel RcW_D. As specified in the following sections, the
closed ring protection tunnel can protect both link failures and node
failures. Wrapping can be applicable for the protection the p2mp
LSPs on the ring, the details of which is outside the scope of this
document.
4.3.1.1. Wrapping for Link Failure
When a link failure between Node B and Node C occurs, if it is a bi-
directional failure, both Node B and Node C can detect the failure
via the OAM mechanism; if it is an uni-directional failure, one of
the two nodes would detect the failure via the OAM mechanism. In
both cases the node at the other side of the detected failure will be
determined by the ring-map and informed using the Ring Protection
Switch Protocol (RPS) which is specified in section 5. Then Node B
switches the clockwise working ring tunnel (RcW_D) to the
anticlockwise protection ring tunnel (RaP_D) and Node C switches
anticlockwise protection ring tunnel(RaP_D) back to the clockwise
working ring tunnel (RcW_D). The payload which enters the ring at
Node A and leaves the ring at Node D follows the path
A->B->A->F->E->D->C->D. The label operation is:
Cheng, et al. Expires December 14, 2017 [Page 11]
Internet-Draft MSRP June 2017
[LSP1](Payload) -> [RcW_D(B)|LSP1](Node A) -> [RaP_D(A)|LSP1](Node B)
-> [RaP_D(F)|LSP1](Node A) -> [RaP_D(E)|LSP1] (Node F) ->
[RaP_D(D)|LSP1] (Node E) -> [RaP_D(C)|LSP1] (Node D) ->
[RcW_D(D)|LSP1](Node C) -> [LSP1](Payload).
+---+#####[RaP_D(F)]######+---+
| F |---------------------| A | +-- LSP1
+---+*****[RcW_D(A)]******+---+
#/* *\#
[RaP_D(E)]#/*[RcW_D(F)] [RcW_D(B)]*\#RaP_D(A)
#/* *\#
+---+ +---+
| E | | B |
+---+ +---+
#\ *x#
[RaP_D(D)]#\ [RcW_D(C)]*x#RaP_D(B)
#\ *x#
+---+*****[RcW_D(D)]****+---+
LSP1 +-- | D |-------------------| C |
+---+#####[RaP_D(C)]####+---+
-----physical links xxxx Failure Link
****** RcW_D ###### RaP_D
Figure 5.Wrapping for link failure
4.3.1.2. Wrapping for Node Failure
As shown in Figure 6, when Node B fails, Node A detects the failure
between A and B and switches the clockwise work ring tunnel (RcW_D)
to the anticlockwise protection ring tunnel (RaP_D), Node C detects
the failure between C and B and switches the anticlockwise protection
ring tunnel (RaP_D) to the clockwise working ring tunnel (RcW_D).
The node at the other side of the failed node will be determined by
the ring-map and informed using the Ring Protection Switch Protocol
(RPS) specified in section 5.
The payload which enters the ring at Node A and exits at Node D
follows the path A->F->E->D->C->D. The label operation is:
[LSP1](Payload)-> [RaP_D(F)|LSP1](NodeA) -> [RaP_D(E)|LSP1](NodeF) ->
[RaP_D(D)|LSP1](NodeE) -> [RaP_D(C)|LSP1] (NodeD) -> [RcW_D(D)|LSP1]
(NodeC) -> [LSP1](Payload).
In one special case where node D fails, all the ring tunnels with
node D as egress will become unusable. The ingress node will update
its ring map according to received RPS messages and determine that
the egress node is not reachable thus it will not send traffic to
Cheng, et al. Expires December 14, 2017 [Page 12]
Internet-Draft MSRP June 2017
either the working or the protection tunnel. However, before the
failure location information is propagated to all the ring nodes, the
wrapping protection mechanism may cause temporary traffic loop: node
C detects the failure and switches the traffic from the clockwise
work ring tunnel (RcW_D) to the anticlockwise protection ring tunnel
(RaP_D), node E also detects the failure and would switch the traffic
from anticlockwise protection ring tunnel (RaP_D) back to the
clockwise work ring tunnel (RcW_D). A possible mechanism to mitigate
the temporary loop problem is: the TTL of the ring tunnel label is
set to 2*N by the ingress ring node of the traffic, where N is the
number of nodes on the ring.
+---+#####[RaP_D(F)]######+---+
| F |---------------------| A | +-- LSP1
+---+*****[RcW_D(A)]******+---+
#/* *\#
[RaP_D(E)]#/*[RcW_D(F)] [RcW_D(B)]*\#RaP_D(A)
#/* *\#
+---+ xxxxx
| E | x B x
+---+ xxxxx
#\ */#
[RaP_D(D)]#\ [RcW_D(C)]*/#RaP_D(B)
#\ */#
+---+*****[RcW_D(D)]****+---+
LSP1 +-- | D |-------------------| C |
+---+#####[RaP_D(C)]####+---+
-----physical links xxxxxx Failure Node
*****RcW_D ###### RaP_D
Figure 6. Wrapping for node failure
4.3.2. Short Wrapping
With the wrapping protection scheme, protection switching is executed
at both nodes adjacent to the failure, consequently the traffic will
be wrapped twice. This mechanism will cause additional latency and
bandwidth consumption when traffic is switched to the protection
path.
With short wrapping protection, protection switching is executed only
at the node upstream to the failure, and the packet leaves the ring
in the protection ring tunnel at the egress node. This scheme can
reduce the additional latency and bandwidth consumption when traffic
is switched to the protection path. However the two directions of a
protected bidirectional LSP are no longer co-routed under the
protection switching conditions.
Cheng, et al. Expires December 14, 2017 [Page 13]
Internet-Draft MSRP June 2017
In the traditional wrapping solution, the protection ring tunnel is
configured as a closed ring, while in the short wrapping solution,
the protection ring tunnel is configured as ended at the egress node,
which is similar to the working ring tunnel. Short wrapping is easy
to implement in shared ring protection because both the working and
protection ring tunnels are terminated on the egress nodes. Figure 7
shows the clockwise working ring tunnel and the anticlockwise
protection ring tunnel with node D as the egress node.
4.3.2.1. Short Wrapping for Link Failure
As shown in Figure 7, in normal state, LSP1 is carried by the
clockwise working ring tunnel (RcW_D) through the path A->B->C->D.
When a link failure between Node B and Node C occurs, Node B switches
the working ring tunnel RcW_D to the protection ring tunnel RaP_D in
the opposite direction. The difference with wrapping occurs in the
protection ring tunnel at the egress node. In short wrapping
protection, Rap_D ends in Node D and then traffic will be forwarded
based on the LSP labels. Thus with the short wrapping mechanism,
LSP1 will follow the path A->B->A->F->E->D when a link failure
between Node B and Node C happens. The protection switch at node D
is based on the information from its ring map and the information
received via the RPS protocol.
+---+#####[RaP_D(F)]######+---+
| F |---------------------| A | +-- LSP1
+---+*****[RcW_D(A)]******+---+
#/* *\#
[RaP_D(E)]#/*[RcW_D(F)] [RcW_D(B)]*\#RaP_D(A)
#/* *\#
+---+ +---+
| E | | B |
+---+ +---+
#\ *x#
[RaP_D(D)]#\ [RcW_D(C)]*x#RaP_D(B)
#\ *x#
+---+*****[RcW_D(D)]****+---+
LSP1 +-- | D |-------------------| C |
+---+ +---+
----- physical links xxxxx Failure Link
****** RcW_D ###### RaP_D
Figure 7. Short wrapping for link failure
Cheng, et al. Expires December 14, 2017 [Page 14]
Internet-Draft MSRP June 2017
4.3.2.2. Short Wrapping for Node Failure
For the node failure which happens on a non-egress node, the short
wrapping protection switching is similar to the link failure case as
described in the previous section. This section specifies the
scenario of an egress node failure.
As shown in Figure 8, LSP1 enters the ring on node A, and leaves the
ring on node D. In normal state, LSP1 is carried by the clockwise
working ring tunnel (RcW_D) through the path A->B->C->D. When node D
fails, traffic of LSP1 cannot be protected by any ring tunnels which
use node D as the egress node. The ingress node will update its ring
map according to received RPS messages and determine that the egress
node is not reachable thus it will not send traffic to either the
working or the protection tunnel. However, before the failure
location information is propagated to all the ring nodes using the
RPS protocol, node C switches all the traffic on the working ring
tunnel RcW_D to the protection ring tunnel RaP_D in the opposite
direction based on the information in the ring map. When the traffic
arrives at node E which also detects the failure of node D, the
protection ring tunnel RaP_D cannot be used to forward traffic to
node D. Since with short wrapping mechanism, protection switching
can only be performed once from the working ring tunnel to the
protection ring tunnel, thus node E MUST NOT switch the traffic which
is already carried on the protection ring tunnel back to the working
ring tunnel in the opposite direction. Instead, node E will discard
the traffic received on RaP_D locally. This can avoid the temporary
traffic loop when the failure happens on the egress node of the ring
tunnel. This also illustrates one of the benefits of having separate
working and protection ring tunnels in each ring direction.
Cheng, et al. Expires December 14, 2017 [Page 15]
Internet-Draft MSRP June 2017
+---+#####[RaP_D(F)]######+---+
| F |---------------------| A | +-- LSP1
+---+*****[RcW_D(A)]******+---+
#/* *\#
[RaP_D(E)]#/*[RcW_D(F)] [RcW_D(B)]*\#RaP_D(A)
#/* *\#
+---+ +---+
| E | | B |
+---+ +---+
#\ */#
[RaP_D(D)]#\ [RcW_D(C)]*/#RaP_D(B)
#\ */#
xxxxx*****[RcW_D(D)]****+---+
LSP1 +-- x D x-------------------| C |
xxxxx +---+
-----physical links xxxxxx Failure Node
*****RcW_D ###### RaP_D
Figure 8. Short Wrapping for egress node failure
4.3.3. Steering
With the steering protection mechanism, the ingress node (which adds
traffic to the ring) perform switching from the working to the
protection ring tunnel, and at the egress node the traffic leaves the
ring from the protection ring tunnel.
When a failure occurs in the ring, the node which detects the failure
with OAM mechanism sends the failure information in the opposite
direction of the failure hop by hop along the ring using an RPS
request message and the ring-map information. When a ring node
receives the RPS message which identifies a failure, it can determine
the location of the fault by using the topology information of the
ring map and updates the ring map accordingly, then it can determine
whether the LSPs entering the ring locally need to switchover or not.
For LSPs that need to switchover, it will switch the LSPs from the
working ring tunnels to their corresponding protection ring tunnels.
4.3.3.1. Steering for Link Failure
Cheng, et al. Expires December 14, 2017 [Page 16]
Internet-Draft MSRP June 2017
Ring map of F +--LSPl
+-+-+-+-+-+-+-+ +---+ ###[RaP_D(F)]### +---/ +-+-+-+-+-+-+-+
|F|A|B|C|D|E|F| | F | ---------------- | A | |A|B|C|D|E|F|A|
+-+-+-+-+-+-+-+ +---+ ***[RcW_D(A)]*** +---+ +-+-+-+-+-+-+-+
|I|I|I|S|I|I| |I|I|S|I|I|I|
+-+-+-+-+-+-+ #/* *\# +-+-+-+-+-+-+
[RaP_D(E)] #/* [RcW_D(B)] *\# [RaP_D(A)]
#/* [RcW_D(F)] *\#
+-+-+-+-+-+-+-+ #/* *\#
|E|F|A|B|C|D|E| +---+ +---+ +-- LSP2
+-+-+-+-+-+-+-+ | E | | B | +-+-+-+-+-+-+-+
|I|I|I|I|S|I| +---+ +---+ |B|C|D|E|F|A|B|
+-+-+-+-+-+-+ #\* */# +-+-+-+-+-+-+-+
#\* [RcW_D(E)] [RcW_D(C)] */# |I|S|I|I|I|I|
[RaP_D(D)] #\* */# +-+-+-+-+-+-+
#\* */# [RaP_D(B)]
+-+-+-+-+-+-+-+ +---+ [RcW_D(D)] +---+ +-+-+-+-+-+-+-+
|D|E|F|A|B|C|D| +-- | D | xxxxxxxxxxxxxxxxx | C | |C|D|E|F|A|B|C|
+-+-+-+-+-+-+-+ LSP1 +---+ [RaP_D(C)] +---+ +-+-+-+-+-+-+-+
|I|I|I|I|I|S| LSP2 |S|I|I|I|I|I|
+-+-+-+-+-+-+ +-+-+-+-+-+-+
----- physical links ***** RcW_D ##### RaP_D
I: Intact S: Severed
Figure 9. Steering operation and protection switching
As shown in Figure 9, LSP1 enters the ring from Node A while LSP2
enters the ring from Node B, and both of them have the same
destination node D.
In normal state, LSP1 is carried by the clockwise working ring tunnel
(RcW_D) through the path A->B->C->D, the label operation is:
[LSP1](Payload) -> [RcW_D(B)|LSP1](NodeA) -> [RcW_D(C)| LSP1](NodeB)
-> [RcW_D(D)|LSP1](NodeC) -> [LSP1](Payload) .
LSP2 is carried by the clockwise working ring tunnel (RcW_D) through
the path B->C->D, the label operation is: [LSP2](Payload) ->
[RcW_D(C)|LSP2](NodeB) -> [RcW_D(D)|LSP2](NodeC) -> [LSP2](Payload) .
If the link between nodes C and D fails, according to the fault
detection and distribution mechanisms, Node D will find out that
there is a failure in the link between C and D, and it will update
the link state of its ring topology, changing the link between C and
D from normal to fault. In the direction that is opposite to the
failure position, Node D will send the state report message to Node
E, informing Node E of the fault between C and D, and E will update
the link state of its ring topology accordingly, changing the link
between C and D from normal to fault. In this way, the state report
Cheng, et al. Expires December 14, 2017 [Page 17]
Internet-Draft MSRP June 2017
message is sent hop by hop in the clockwise direction. Similar to
Node D, Node C will send the failure information in the anti-
clockwise direction.
When Node A receives the failure report message and updates the link
state of its ring map, it is aware that there is a fault on the
clockwise working ring tunnel to node D (RcW_D), and LSP1 enters the
ring locally and is carried by this ring tunnel, thus Node A will
decide to switch the LSP1 onto the anticlockwise protection ring
tunnel to node D (RaP_D). After the switchover, LSP1 will follow the
path A->F->E->D, the label operation is: [LSP1](Payload) ->
[RaP_D(F)| LSP1](NodeA) -> [RaP_D(E)|LSP1](NodeF) ->
[RaP_D(D)|LSP1](NodeE) -> [LSP1](Payload).
The same procedure also applies to the operation of LSP2. When Node
B updates the link state of its ring topology, and finds out that the
working ring tunnel RcW_D has failed, it will switch the LSP2 to the
anticlockwise protection tunnel RaP_D. After the switchover, LSP2
goes through the path B->A->F->E->D, and the label operation is:
[LSP2](Payload) -> [RaP_D(A)|LSP2](NodeB) -> [RaP_D(F)|LSP2](NodeA)
-> [RaP_D(E)|LSP2](NodeF) -> [RaP_D(D)|LSP2](NodeE) ->
[LSP2](Payload).
Assume the link between nodes A and B breaks down, as shown in
Figure 10. Similar to the above failure case, Node B will detect a
fault in the link between A and B, and it will update its ring map,
changing the link state between A and B from normal to fault. The
state report message is sent hop by hop in the clockwise direction,
notifying every node that there is a fault between node A and B, and
every node updates the link state of its ring topology. As a result,
Node A will detect a fault in the working ring tunnel to node D, and
switch LSP1 to the protection ring tunnel, while Node B determine
that the working ring tunnel for LSP2 still works fine, and will not
perform the switchover.
Cheng, et al. Expires December 14, 2017 [Page 18]
Internet-Draft MSRP June 2017
/-- LSPl
+-+-+-+-+-+-+-+ +---+ ###[RaP_D(F)]#### +---/ +-+-+-+-+-+-+-+
|F|A|B|C|D|E|F| | F | ----------------- | A | |A|B|C|D|E|F|A|
+-+-+-+-+-+-+-+ +---+ ***[RcW_D(A)]**** +---+ +-+-+-+-+-+-+-+
|I|S|I|I|I|I| #/* x |S|I|I|I|I|I|
+-+-+-+-+-+-+ #/* x +-+-+-+-+-+-+
[RaP_D(E)] #/*[RcW_D(F)] [RcW_D(B)]x [RaP_D(A)]
#/* x +-- LSP2
+-+-+-+-+-+-+-+ +---+ +---++-+-+-+-+-+-+-+
|E|F|A|B|C|D|E| | E | | B ||B|C|D|E|F|A|B|
+-+-+-+-+-+-+-+ +---+ +---++-+-+-+-+-+-+-+
|I|I|S|I|I|I| #\* */# |I|I|I|I|I|S|
+-+-+-+-+-+-+ #\*[RcW_D(E)] [RcW_D(C)] */# +-+-+-+-+-+-+
[RaP_D(D)] #\* */# [RaP_D(B)]
+-+-+-+-+-+-+-+ #\* */# +-+-+-+-+-+-+-+
|D|E|F|A|B|C|D| +---+ ***[RcW_D(D)]*** +---+ |C|D|E|F|A|B|C|
+-+-+-+-+-+-+-+ +-- | D | ---------------- | C | +-+-+-+-+-+-+-+
|I|I|I|S|I|I| LSP1 +---+ ###[RaP_D(C)]### +---+ |I|I|I|I|S|I|
+-+-+-+-+-+-+ LSP2 +-+-+-+-+-+-+
----- physical links ***** RcW_D ##### RaP_D
Figure 10. Steering operation and protection switching (2)
4.3.3.2. Steering for Node Failure
For a node failure which happens on a non-egress node, steering
protection switching is similar to the link failure case as described
in the previous section.
If the failure occurs at the egress node of the LSP, the ingress node
will update its ring map according to the received RPS messages, it
will also determine that the egress node is not reachable after the
failure, thus it will not send traffic to either the working or the
protection tunnel, and a traffic loop can be avoided.
4.4. Interconnected Ring Protection
4.4.1. Interconnected Ring Topology
Interconnected ring topology is widely used in MPLS-TP networks. For
a given ring, the interconnection node acts as the egress node for
that ring, meaning that all LSPs using the interconnection node as an
egress from one specific ring to another will use the same group of
ring tunnels within the ring. This document will discuss two typical
interconnected ring topologies:
1. Single-node interconnected rings
Cheng, et al. Expires December 14, 2017 [Page 19]
Internet-Draft MSRP June 2017
In single-node interconnected rings, the connection between
the two rings is through a single node. Because the
interconnection node is in fact a single point of failure,
this topology should be avoided in real transport networks.
Figure 11 shows the topology of single-node interconnected
rings. Node C is the interconnection node between Ring1 and
Ring2.
+---+ +---+ +---+ +---+
| A |------| B |----- -----| G |------| H |
+---+ +---+ \ / +---+ +---+
| \ / |
| \ +---+ / |
| Ring1 | C | Ring2 |
| / +---+ \ |
| / \ |
+---+ +---+ / \ +---+ +---+
| F |------| E |----- -----| J |------| I |
+---+ +---+ +---+ +---+
Figure 11. Single-node interconnected rings
2. Dual-node interconnected rings
In dual-node interconnected rings, the connection between the
two rings is through two nodes. The two interconnection nodes
belong to both interconnected rings. This topology can
recover from one interconnection node failure.
Figure 12 shows the topology of dual-node interconnected
rings. Nodes C and Node D are the interconnection nodes
between Ring1 and Ring2.
Cheng, et al. Expires December 14, 2017 [Page 20]
Internet-Draft MSRP June 2017
+---+ +---+ +---+ +---+ +---+
| A |------| B |------| C |------| G |------| H |
+---+ +---+ +---+ +---+ +---+
| | | |
| | | |
| Ring1 | | Ring2 |
| | | |
| | | |
+---+ +---+ +---+ +---+ +---+
| F |------| E |------| D |------| J |------| I |
+---+ +---+ +---+ +---+ +---+
Figure 12. Dual-node interconnected rings
4.4.2. Interconnected Ring Protection Mechanisms
Interconnected rings can be treated as two independent rings. Ring
protection switching (RPS) protocol operates on each ring
independently. A failure that happens in one ring only triggers
protection switching in the ring itself and does not affect the other
ring, unless the failure is on the interconnection node. In this
way, protection switching on each ring is the same as the mechanisms
described in section 4.3.
The service LSPs that traverse the interconnected rings use the ring
tunnels in each ring, within a given ring, the tunnel is selected
using normal ring selection procedures. The traversing LSPs are
stitched on the interconnection node. On the interconnection node,
the ring tunnel label of the source ring is popped, then LSP label is
swapped, after that the ring tunnel label of the destination ring is
pushed.
In the dual-node interconnected ring scenario, the two
interconnection nodes can be managed as a virtual node group. In
addition to the ring tunnels to each physical ring node, Each ring
SHOULD assign the working and protection ring tunnels to the virtual
interconnection node group. In addition, on both nodes in the
virtual interconnection node group, the same LSP label is assigned
for each traversed LSP. This way, any interconnection node in the
virtual node group can terminate the working or protection ring
tunnels targeted to the virtual node group, and stitch the service
LSP from the source ring tunnel to the destination ring tunnel.
When the service LSP passes through the interconnected rings, the
direction of the working ring tunnels used on both rings SHOULD be
the same. In dual-node interconnected rings, this ensures that in
normal state the traffic passes only one of the two interconnection
nodes, and does not pass the link between the two interconnection
Cheng, et al. Expires December 14, 2017 [Page 21]
Internet-Draft MSRP June 2017
nodes. The traffic will then only be switched to the protection path
if the interconnection node which is in working path fails. For
example, if the service LSP uses the clockwise working ring tunnel on
Ring1, when the service LSP leaves Ring1 and enters Ring2, the
working ring tunnel used on Ring2 should also follow the clockwise
direction.
4.4.3. Ring Tunnels in Interconnected Rings
The same ring tunnels as described in section 4.1 are used in each
ring of the interconnected rings. In addition, ring tunnels to the
virtual interconnection node group are established on each ring of
the interconnected rings, i.e.:
o one clockwise working ring tunnel to the virtual interconnection
node group
o one anticlockwise protection ring tunnel to the virtual
interconnection node group
o one anticlockwise working ring tunnel to the virtual
interconnection node group
o one clockwise protection ring tunnel to the virtual
interconnection node group
The ring tunnels to the virtual interconnection node group are shared
by all LSPs that need to be forwarded to other rings. These ring
tunnels can terminate at any node in the virtual interconnection node
group.
For example, all the ring tunnels on Ring1 in Figure 13 are
provisioned as follows:
o To Node A: R1cW_A, R1aW_A, R1cP_A, R1aP_A
o To Node B: R1cW_B, R1aW_B, R1cP_B, R1aP_B
o To Node C: R1cW_C, R1aW_C, R1cP_C, R1aP_C
o To Node D: R1cW_D, R1aW_D, R1cP_D, R1aP_D
o To Node E: R1cW_E, R1aW_E, R1cP_E, R1aP_E
o To Node F: R1cW_F, R1aW_F, R1cP_F, R1aP_F
o To the virtual interconnection node group (including Node F and
Node A): R1cW_F&A, R1aW_F&A, R1cP_F&A, R1aP_F&A
Cheng, et al. Expires December 14, 2017 [Page 22]
Internet-Draft MSRP June 2017
All the ring tunnels on Ring2 in Figure 13 are provisioned as
follows:
o To Node A: R2cW_A, R2aW_A, R2cP_A, R2aP_A
o To Node F: R2cW_F, R2aW_F, R2cP_F, R2aP_F
o To Node G: R2cW_G, R2aW_G, R2cP_G, R2aP_G
o To Node H: R2cW_H, R2aW_H, R2cP_H, R2aP_H
o To Node I: R2cW_I, R2aW_I, R2cP_I, R2aP_I
o To Node J: R2cW_J, R2aW_J, R2cP_J, R2aP_J
o To the virtual interconnection node group (including Node F and
Node A): R2cW_F&A, R2aW_F&A, R2cP_F&A, R2aP_F&A
Cheng, et al. Expires December 14, 2017 [Page 23]
Internet-Draft MSRP June 2017
+---+cccccccccccc +---+
| H |-------------| I |--->LSP1
+---+ +---+
c/a a\
c/a a\
c/a a\
+---+ +---+
| G | Ring2 | J |
+---+ +---+
c\a a/c
c\a a/c
c\a aaaaaaaaaaaaa a/c
+---+ccccccccccccc+---+
| F |-------------| A |
+---+ccccccccccccc+---+
c/aaaaaaaaaaaaaaaaaaa a\
c/ a\
c/ a\
+---+ +---+
| E | Ring1 | B |
+---+ +---+
c\a a/c
c\a a/c
c\a a/c
+---+aaaaaaaaaaaa +---+
LSP1--->| D |-------------| C |
+---+ccccccccccccc+---+
ccccccccccc R1cW_F&A
aaaaaaaaaaa R1aP_F&A
ccccccccccc R2cW_I
aaaaaaaaaaa R2aP_I
Figure 13. Ring tunnels for the interconnected rings
4.4.4. Interconnected Ring Switching Procedure
As shown in Figure 13, for the service LSP1 which enters Ring1 at
Node D and leaves Ring1 at Node F and continues to enter Ring2 at
Node F and leaves Ring2 at Node I, the short wrapping protection
scheme is described as below.
In normal state, LSP1 follows R1cW_F&A in Ring1 and R2cW_I in Ring2.
At the interconnection node F, the label used for the working ring
tunnel R1cW_F&A in Ring1 is popped, the LSP label is swapped, and the
label used for the working ring tunnel R2cW_I in Ring2 will be pushed
based the inner LSP label lookup. The working path that the service
LSP1 follows is: LSP1->R1cW_F&A (D->E->F)->R2cW_I(F->G->H->I)->LSP1.
Cheng, et al. Expires December 14, 2017 [Page 24]
Internet-Draft MSRP June 2017
In case of link failure, for example, when a failure occurs on the
link between Node F and Node E, Node E will detect the failure and
execute protection switching as described in 4.3.2. The path that
the service LSP1 follows after switching change to: LSP1->R1cW_F&A(D-
>E)->R1aP_F&A(E->D->C->B->A)->R2cW_I(A->F->G->H->I)->LSP1.
In case of a non-interconnection node failure, for example, when the
failure occurs at Node E in Ring1, Node D will detect the failure and
execute protection switching as described in 4.3.2. The path that
the service LSP1 follows after switching becomes:
LSP1->R1aP_F&A(D->C->B->A)->R2cW_I(A->F->G->H->I)->LSP1.
In case of an interconnection node failure, for example, when the
failure occurs at the interconnection Node F, Node E in Ring1 will
detect the failure, and execute protection switching as described in
4.3.2. Node A in Ring2 will also detect the failure, and execute
protection switching as described in 4.3.2. The path that the
service traffic LSP1 follows after switching is:
LSP1->R1cW_F&A(D->E)->R1aP_F&A(E->D->C->B->A)->R2aP_I(A->J->I)->LSP1.
4.4.5. Interconnected Ring Detection Mechanism
As shown in Figure 13, in normal state the service traffic LSP1
traverses D->E->F in Ring1 and F->G->H->I in Ring2. Node A and F are
the interconnection nodes. When both the link between Node F and
Node G and the link between Node F and Node A fail, the ring tunnel
from Node F to Node I in Ring2 becomes unreachable. However, the
other interconnection node A is still available, and LSP1 can still
reach Node I via node A.
In order to achieve this, the interconnection nodes need to know the
ring topology of each ring so that they can judge whether a node is
reachable. This judgment is based on the knowledge of ring map and
the fault location. The ring map can be obtained from the NMS or
topology discovery mechanisms. The fault location can be obtained by
transmitting the fault information around the ring. The nodes that
detect the failure will transmit the fault information in the
opposite direction hop by hop using the RPS protocol message. When
the interconnection node receives the message that informs the
failure, it will calculate the location of the fault according to the
topology information that is maintained by itself and determines
whether the LSPs entering the ring at itself can reach the
destination. If the destination node is reachable, the LSP will
leave the source ring and enter the destination ring. If the
destination node is not reachable, the LSP will switch to the
anticlockwise protection ring tunnel.
Cheng, et al. Expires December 14, 2017 [Page 25]
Internet-Draft MSRP June 2017
In Figure 13, Node F determines that the ring tunnel to Node I is
unreachable, the service LSP1 for which the destination node on the
ring2 is Node I MUST switch to the protection ring tunnel (R1aP_F&A)
and consequently the service traffic LSP1 traverses the
interconnected rings at Node A. Node A will pop the ring tunnel
label of Ring1 and push the ring tunnel label of Ring2 and send the
traffic to Node I via ring tunnel (R2aW_I).
5. Ring Protection Coordination Protocol
5.1. RPS and PSC Comparison on Ring Topology
This section provides comparison between RPS and PSC [RFC6378]
[RFC6974] on ring topologies. This can be helpful to explain the
reason of defining a new protocol for ring protection switching.
The PSC protocol [RFC6378] is designed for point-to-point LSPs, on
which the protection switching can only be performed on one or both
of the end points of the LSP. The RPS protocol is designed for ring
tunnels, which consist of multiple ring nodes, and the failure could
happen on any segment of the ring, thus RPS is capable of identifying
and handling the different failures on the ring, and coordinating the
protection switching behavior of all the nodes on the ring. As will
be specified in the following sections, this is achieved with the
introduction of the "Pass-Through" state for the ring nodes, and the
location of the protection request is identified via the Node IDs in
the RPS Request message.
Taking a ring topology with N nodes as example:
With the mechanism specified in [RFC6974], on every ring node, a
linear protection configuration has to be provisioned with every
other node in the ring, i.e. with (N-1) other nodes. This means that
on every ring node there will be (N-1) instances of the PSC protocol.
And in order to detect faults and to transport the PSC message, each
instance shall have a MEP on the working path and a MEP on the
protection path respectively. This means that every node on the ring
needs to be configured with (N-1) * 2 MEPs.
With the mechanism defined in this document, on every ring node there
will only be a single instance of the RPS protocol. In order to
detect faults and to transport the RPS message, each node only needs
to have a MEP on the section to its adjacent nodes respectively. In
this way, every ring node only needs to be configured with 2 MEPs.
As shown in the above example, RPS is designed for ring topologies
and can achieve ring protection efficiently with minimum protection
Cheng, et al. Expires December 14, 2017 [Page 26]
Internet-Draft MSRP June 2017
instances and OAM entities, which meets the requirements on topology
specific recovery mechanisms as specified in [RFC5654].
5.2. RPS Protocol
The Ring Protection Switch (RPS) Protocol defined in this section is
used to coordinate the protection switching action of all the ring
nodes in the same ring.
The protection operation of the ring tunnels is controlled with the
help of the RPS protocol. The RPS processes in each of the
individual ring nodes that form the ring MUST communicate using the
G-ACh channel. The RPS protocol is applicable to all the three ring
protection modes. This section takes the short-wrapping mechanism
described in section 4.3.2 as an example.
The RPS protocol is used to distribute the ring status information
and RPS requests to all the ring nodes. Changes in the ring status
information and RPS requests can be initiated automatically based on
link status or caused by external commands.
Each node on the ring is uniquely identified by assigning it a node
ID. The node ID MUST be unique on each ring. The maximum number of
nodes on the ring supported by the RPS protocol is 127. The node ID
SHOULD be independent of the order in which the nodes appear on the
ring. The node ID is used to identity the source and destination
nodes of each RPS request.
Every node obtains the ring topology either by configuration or via
some topology discovery mechanism. The ring map consists of the ring
topology information, and connectivity status (Intact or Severed)
between the adjacent ring nodes, which is determined via the OAM
message exchanged between the adjacent nodes. The ring map is used
by every ring node to determine the switchover behavior of the ring
tunnels.
As shown in Figure 14, when no protection switching is active on the
ring, each node MUST send RPS requests with No Request (NR) to its
two adjacent nodes periodically. The transmission interval of RPS
requests is specified in section 5.2.1.
+---+ A->B(NR) +---+ B->C(NR) +---+ C->D(NR)
-------| A |-------------| B |-------------| C |-------
(NR)F<-A +---+ (NR)A<-B +---+ (NR)B<-C +---+
Figure 14. RPS communication between the ring nodes in case of
no failure in the ring
Cheng, et al. Expires December 14, 2017 [Page 27]
Internet-Draft MSRP June 2017
As shown in Figure 15, When a node detects a failure and determines
that protection switching is required, it MUST send the appropriate
RPS request in both directions to the destination node. The
destination node is the other node that is adjacent to the identified
failure. When a node that is not the destination node receives an
RPS request and it has no higher priority local request, it MUST
transfer in the same direction the RPS request as received. In this
way, the switching nodes can maintain RPS protocol communication in
the ring. The RPS request MUST be terminated by the destination node
of the message. If an RPS request with the node itself set as the
source node is received, this message MUST be dropped and not be
forwarded to next node.
+---+ C->B(SF) +---+ B->C(SF) +---+ C->B(SF)
-------| A |-------------| B |----- X -----| C |-------
(SF)C<-B +---+ (SF)C<-B +---+ (SF)B<-C +---+
Figure 15. RPS communication between the ring nodes in case of
failure between nodes B and C
Note that in the case of a bidirectional failure such as a cable cut,
the two adjacent nodes detect the failure and send each other an RPS
request in opposite directions.
o In rings utilizing the wrapping protection, each node detects the
failure or receives the RPS request as the destination node MUST
perform the switch from/to the working ring tunnels to/from the
protection ring tunnels if it has no higher priority active RPS
request.
o In rings utilizing the short wrapping protection, each node
detects the failure or receives the RPS request as the destination
node MUST perform the switch only from the working ring tunnels to
the protection ring tunnels.
o In rings utilizing the steering protection. When a ring switch is
required, any node MUST perform the switches if its added/dropped
traffic is affected by the failure. Determination of the affected
traffic MUST be performed by examining the RPS requests
(indicating the nodes adjacent to the failure or failures) and the
stored ring map (indicating the relative position of the failure
and the added traffic destined towards that failure).
When the failure has cleared and the Wait-to-Restore (WTR) timer has
expired, the nodes which generate the RPS requests MUST drop their
respective switches and MUST generate an RPS request carrying the NR
code. The node receiving from both directions such an RPS request
MUST drop its protection switches.
Cheng, et al. Expires December 14, 2017 [Page 28]
Internet-Draft MSRP June 2017
A protection switch MUST be initiated by one of the criteria
specified in Section 5.3. A failure of the RPS protocol or
controller MUST NOT trigger a protection switch.
Ring switches MUST be preempted by higher priority RPS requests. For
example, consider a protection switch that is active due to a manual
switch request on the given link, and another protection switch is
required due to a failure on another link. Then an RPS request MUST
be generated, the former protection switch MUST be dropped, and the
latter protection switch established.
The shared ring protection mechanism supports multiple protection
switches in the ring, resulting in the ring being segmented into two
or more separate segments. This may happen when several RPS requests
of the same priority exist in the ring due to multiple failures or
external switch commands.
Proper operation of the MSRP mechanism relies on all nodes using
their ring map to determine the state of the ring (nodes and links).
In order to accommodate ring state knowledge, during a protection
switch the RPS requests MUST be sent in both directions.
5.2.1. Transmission and Acceptance of RPS Requests
A new RPS request MUST be transmitted immediately when a change in
the transmitted status occurs.
The first three RPS protocol messages carrying new RPS request MUST
be transmitted as fast as possible. For fast protection switching
within 50 ms, the interval of the first three RPS protocol messages
SHOULD be 3.3 ms. The successive RPS requests SHOULD be transmitted
with the interval of 5 seconds. A ring node which is not the
destination of the received RPS message MUST forward it to the next
node along the ring immediately.
5.2.2. RPS PDU Format
Figure 16 depicts the format of an RPS packet that is sent on the
G-ACh. The Channel Type field is set to indicate that the message is
an RPS message.
Cheng, et al. Expires December 14, 2017 [Page 29]
Internet-Draft MSRP June 2017
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|0 0 0 1|Version| Reserved | RPS Channel Type (TBD) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Dest Node ID | Src Node ID | Request | M | Reserved |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 16. G-ACh RPS Packet Format
The following fields MUST be provided:
o Destination Node ID: The destination node ID MUST always be set to
value of the node ID of the adjacent node. The Node ID MUST be
unique on each ring. Valid destination node ID values are 1-127.
o Source Node ID: The source node ID MUST always be set to the ID
value of the node generating the RPS request. The Node ID MUST be
unique on each ring. Valid source node ID values are 1-127.
o Protection Switching Mode (M): This 2-bit field indicates the
protection switching mode used by the sending node of the RPS
message. This can be used to check that the ring nodes on the
same ring use the same protection switching mechanism. The
defined values of the M field are listed as below:
+------------------+-----------------------------+
| Bits (MSB-LSB) | Protecton Switching Mode |
+------------------+-----------------------------+
| 0 0 | Reserved |
| 0 1 | Wrapping |
| 1 0 | Short Wrapping |
| 1 1 | Steering |
+------------------+-----------------------------+
o RPS request code: A code consisting of eight bits as specified
below:
Cheng, et al. Expires December 14, 2017 [Page 30]
Internet-Draft MSRP June 2017
+------------------+-----------------------------+----------+
| Bits | Condition, State | Priority |
| (MSB - LSB) | or external Request | |
+------------------+-----------------------------+----------+
| 0 0 0 0 1 1 1 1 | Lockout of Protection (LP) | highest |
| 0 0 0 0 1 1 0 1 | Forced Switch (FS) | |
| 0 0 0 0 1 0 1 1 | Signal Fail (SF) | |
| 0 0 0 0 0 1 1 0 | Manual Switch (MS) | |
| 0 0 0 0 0 1 0 1 | Wait-To-Restore (WTR) | |
| 0 0 0 0 0 0 1 1 | Exercise (EXER) | |
| 0 0 0 0 0 0 0 1 | Reverse Request (RR) | |
| 0 0 0 0 0 0 0 0 | No Request (NR) | lowest |
+------------------+-----------------------------+----------+
5.2.3. Ring Node RPS States
Idle state: A node is in the idle state when it has no RPS request
and is sending and receiving NR code to/from both directions.
Switching state: A node not in the idle or pass-through states is in
the switching state.
Pass-through state: A node is in the pass-through state when its
highest priority RPS request is a request not destined to it or
generated by it. The pass-through is bidirectional.
5.2.3.1. Idle State
A node in the idle state MUST generate the NR request in both
directions.
A node in the idle state MUST terminate RPS requests flow in both
directions.
A node in the idle state MUST block the traffic flow on protection
ring tunnels in both directions.
5.2.3.2. Switching State
A node in the switching state MUST generate RPS request to its
adjacent node with its highest RPS request code in both directions
when it detects a failure or receives an external command.
In bidirectional failure condition, both of the nodes adjacent to the
failure detect the failure and send the RPS request in both
directions with the destination set to each other, while each node
can only receive the RPS request via the long path, the message sent
via the short path will get lost due to the bidirectional failure.
Cheng, et al. Expires December 14, 2017 [Page 31]
Internet-Draft MSRP June 2017
Here the short path refers to the shorter path on the ring between
the source and destination node of the RPS request, and the long path
refers to the longer path on the ring between the source and
destination node of the RPS request. Upon receipt of the RPS request
on the long path, the destination node of the RPS request MUST send
RPS request with its highest request code periodically along the long
path to the other node adjacent to the failure.
In unidirectional failure condition, the node which detects the
failure MUST send the RPS request in both directions with the
destination node set to the other node adjacent to the failure. The
destination node of the RPS request cannot detect the failure itself
but will receive RPS request from both the short path and the long
path. The destination node MUST acknowledge the received RPS request
by replying an RPS request with the RR code on the short path, and an
RPS request with the received RPS request code on the long path.
Accordingly, when the node which detects the failure receives RPS
request with RR code on the short path, then the RPS request received
from the same node along the long path SHOULD be ignored.
A node in the switching state MUST terminate the received RPS
requests in both directions and not forward it further along the
ring.
The following switches as defined in section 5.3.1 MUST be allowed to
coexist:
o LP and LP
o FS and FS
o SF and SF
o FS and SF
When multiple MS RPS requests exist at the same time addressing
different links and there is no higher priority request on the ring,
no switch SHOULD be executed and existing switches MUST be dropped.
The nodes MUST still signal RPS request with the MS code.
Multiple EXER requests MUST be allowed to coexist in the ring.
A node in a ring switching state that receives the external command
LP for the affected link MUST drop its switch and MUST signal NR for
the locked link if there is no other RPS request on another link.
The node still SHOULD signal relevant RPS request for another link.
Cheng, et al. Expires December 14, 2017 [Page 32]
Internet-Draft MSRP June 2017
5.2.3.3. Pass-through State
When a node is in a pass-through state, it MUST transfer the received
RPS Request unchanged in the same direction.
When a node is in a pass-through state, it MUST enable the traffic
flow on protection ring tunnels in both directions.
5.2.4. RPS State Transitions
All state transitions are triggered by an incoming RPS request
change, a WTR expiration, an externally initiated command, or locally
detected MPLS-TP section failure conditions.
RPS requests due to a locally detected failure, an externally
initiated command, or received RPS request shall preempt existing RPS
requests in the prioritized order given in Section 5.2.2, unless the
requests are allowed to coexist.
5.2.4.1. Transitions Between Idle and Pass-through States
The transition from the idle state to pass-through state MUST be
triggered by a valid RPS request change, in any direction, from the
NR code to any other code, as long as the new request is not destined
to the node itself. Both directions move then into a pass-through
state, so that, traffic entering the node through the protection ring
tunnels are transferred transparently through the node.
A node MUST revert from pass-through state to the idle state when RPS
request with NR code is received in both directions. Then both
directions revert simultaneously from the pass-through state to the
idle state.
5.2.4.2. Transitions Between Idle and Switching States
Transition of a node from the Idle state to the Switching state MUST
be triggered by one of the following conditions:
o A valid RPS request change from the NR code to any code received
on either the long or the short path and is destined to this node
o An externally initiated command for this node
o The detection of an MPLS-TP section layer failure at this node
Actions taken at a node in the Idle state upon transition to the
switching state are:
Cheng, et al. Expires December 14, 2017 [Page 33]
Internet-Draft MSRP June 2017
o For all protection switch requests, except EXER and LP, the node
MUST execute the switch
o For EXER, and LP, the node MUST signal appropriate request but not
execute the switch
In one of the following conditions, transition from the Switching
state to the Idle state MUST be triggered:
o On node which triggers the protection switching, when the WTR time
expires or an externally initiated command is cleared, the node
MUST transit from Switching state to Idle State and signal the NR
code using RPS message in both directions.
o On node which enters the switching state due to the received RPS
request: Upon reception of the NR code from both directions, the
head-end node MUST drop its switch, transition to Idle State and
signal the NR code in both directions.
5.2.4.3. Transitions Between Switching States
When a node that is currently executing any protection switch
receives a higher priority RPS request (due to a locally detected
failure, an externally initiated command, or a ring protection switch
request destined to it) for the same link, it MUST update the
priority of the switch it is executing to the priority of the
received RPS request.
When a failure condition clears at a node, the node MUST enter WTR
condition and remain in it for the appropriate time-out interval,
unless:
o A different RPS request with a higher priority than WTR is
received
o Another failure is detected
o An externally initiated command becomes active
The node MUST send out a WTR code on both the long and short paths.
When a node that is executing a switch in response to incoming SF RPS
request (not due to a locally detected failure) receives a WTR code
(unidirectional failure case), it MUST send out RR code on the short
path and the WTR on the long path.
Cheng, et al. Expires December 14, 2017 [Page 34]
Internet-Draft MSRP June 2017
5.2.4.4. Transitions Between Switching and Pass-through States
When a node that is currently executing a switch receives an RPS
request for a non-adjacent link of higher priority than the switch it
is executing, it MUST drop its switch immediately and enter the pass-
through state.
The transition of a node from pass-through to switching state MUST be
triggered by:
o An equal priority, a higher priority, or an allowed coexisting
externally initiated command
o The detection of an equal priority, a higher priority, or an
allowed coexisting automatic initiated command
o The receipt of an equal, a higher priority, or an allowed
coexisting RPS request destined to this node
5.3. RPS State Machine
5.3.1. Switch Initiation Criteria
5.3.1.1. Administrative Commands
Administrative commands can be initiated by the network operator
through the Network Management System (NMS). The operator command
may be transmitted to the appropriate node via the MPLS-TP RPS
message.
The following commands can be transferred by the RPS message:
o Lockout of Protection (LP): This command prevents any protection
activity and prevents using ring switches anywhere in the ring.
If any ring switches exist in the ring, this command causes the
switches to drop.
o Forced Switch to protection (FS): This command performs the ring
switch of normal traffic from the working entity to the protection
entity for the link between the node at which the command is
initiated and the adjacent node to which the command is directed.
This switch occurs regardless of the state of the MPLS-TP section
for the requested link, unless a higher priority switch request
exists.
o Manual Switch to protection (MS): This command performs the ring
switch of the normal traffic from the working entity to the
protection entity for the link between the node at which the
Cheng, et al. Expires December 14, 2017 [Page 35]
Internet-Draft MSRP June 2017
command is initiated and the adjacent node to which the command is
directed. This occurs if the MPLS-TP section for the requested
link is not satisfying an equal or higher priority switch request.
o Exercise - Ring (EXER): This command exercises ring protection
switching on the addressed link without completing the actual
switch. The command is issued and the responses (RR) are checked,
but no normal traffic is affected.
The following commands are not transferred by the RPS message:
o Clear: This command clears the administrative command and Wait-To-
Restore timer (WTR) at the node to which the command was
addressed. The node to node signaling after the removal of the
externally initiated commands is performed using the no-request
code (NR).
o Lockout of Working (LW): This command prevents the normal traffic
transported over the addressed link from being switched to the
protection entity by disabling the node's capability of requesting
switch for this link in case of failure. If any normal traffic is
already switched on the protection entity, the switch is dropped.
If no other switch requests are active on the ring, the no-request
code (NR) is transmitted. This command has no impact on any other
link. If the node receives the switch request from the adjacent
node from any side it will perform the requested switch. If the
node receives the switch request addressed to the other node, it
will enter the pass-through state.
5.3.1.2. Automatically Initiated Commands
Automatically initiated commands can be initiated based on MPLS-TP
section layer OAM indication and the received switch requests.
The node can initiate the following switch requests automatically:
o Signal Fail (SF): This command is issued when the MPLS-TP section
layer OAM detects signal failure condition.
o Wait-To-Restore (WTR): This command is issued when MPLS-TP section
detects that the SF condition has cleared. It is used to maintain
the state during the WTR period unless it is preempted by a higher
priority switch request. The WTR time may be configured by the
operator in 1 minute steps between 0 and 12 minutes; the default
value is 5 minutes.
Cheng, et al. Expires December 14, 2017 [Page 36]
Internet-Draft MSRP June 2017
o Reverse Request (RR): This command is transmitted to the source
node of the received RPS message over the short path as an
acknowledgment for receiving the switch request.
5.3.2. Initial States
This section describes the possible states of a ring node, the
corresponding action of the working and protection ring tunnels on
the node, and the RPS request which should be generated in that
state.
Cheng, et al. Expires December 14, 2017 [Page 37]
Internet-Draft MSRP June 2017
+-----------------------------------+----------------+
| State | Signaled RPS |
+-----------------------------------+----------------+
| A | Idle | NR |
| | Working: no switch | |
| | Protection: no switch | |
+-----+-----------------------------+----------------+
| B | Pass-through | N/A |
| | Working: no switch | |
| | Protection: pass through | |
+-----+-----------------------------+----------------+
| C | Switching - LP | LP |
| | Working: no switch | |
| | Protection: no switch | |
+-----+-----------------------------+----------------+
| D | Idle - LW | NR |
| | Working: no switch | |
| | Protection: no switch | |
+-----+-----------------------------+----------------+
| E | Switching - FS | FS |
| | Working: switched | |
| | Protection: switched | |
+-----+-----------------------------+----------------+
| F | Switching - SF | SF |
| | Working: switched | |
| | Protection: switched | |
+-----+-----------------------------+----------------+
| G | Switching - MS | MS |
| | Working: switched | |
| | Protection: switched | |
+-----+-----------------------------+----------------+
| H | Switching - WTR | WTR |
| | Working: switched | |
| | Protection: switched | |
+-----+-----------------------------+----------------+
| I | Switching - EXER | EXER |
| | Working: no switch | |
| | Protection: no switch | |
+-----+-----------------------------+----------------+
5.3.3. State transitions When Local Request is Applied
In the state description below 'O' means that new local request will
be rejected because of exiting request.
=====================================================================
Initial state New request New state
------------- ----------- ---------
Cheng, et al. Expires December 14, 2017 [Page 38]
Internet-Draft MSRP June 2017
A (Idle) LP C (Switching - LP)
LW D (Idle - LW)
FS E (Switching - FS)
SF F (Switching - SF)
Recover from SF N/A
MS G (Switching - MS)
Clear N/A
WTR expires N/A
EXER I (Switching - EXER)
=====================================================================
Initial state New request New state
------------- ----------- ---------
B (Pass-through) LP C (Switching - LP)
LW B (Pass-through)
FS O - if current state is due to
LP sent by another node
E (Switching - FS) - otherwise
SF O - if current state is due to
LP sent by another node
F (Switching - SF) - otherwise
Recover from SF N/A
MS O - if current state is due to
LP, SF or FS sent by
another node
G (Switching - MS) - otherwise
Clear N/A
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
C (Switching - LP) LP N/A
LW O
FS O
SF O
Recover from SF N/A
MS O
Clear A (Idle) - if there is no
failure in the ring
F (Switching - SF) - if there
is a failure at this node
B (Pass-through) - if there is
a failure at another node
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
Cheng, et al. Expires December 14, 2017 [Page 39]
Internet-Draft MSRP June 2017
D (Idle - LW) LP C (Switching - LP)
LW N/A - if on the same link
D (Idle - LW) - if on another
link
FS O - if on the same link
E (Switching - FS) - if on
another link
SF O - if on the addressed link
F (Switching - SF) - if on
another link
Recover from SF N/A
MS O - if on the same link
G (Switching - MS) - if on
another link
Clear A (Idle) - if there is no
failure on addressed link
F (Switching - SF) - if there
is a failure on this link
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
E (Switching - FS) LP C (Switching - LP)
LW O - if on another link
D (Idle - LW) - if on the same
link
FS N/A - if on the same link
E (Switching - FS) - if on
another link
SF O - if on the addressed link
E (Switching - FS) - if on
another link
Recover from SF N/A
MS O
Clear A (Idle) - if there is no
failure in the ring
F (Switching - SF) - if there
is a failure at this node
B (Pass-through) - if there is
a failure at another node
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
F (Switching - SF) LP C (Switching - LP)
LW O - if on another link
Cheng, et al. Expires December 14, 2017 [Page 40]
Internet-Draft MSRP June 2017
D (Idle - LW) - if on the same
link
FS E (Switching - FS)
SF N/A - if on the same link
F (Switching - SF) - if on
another link
Recover from SF H (Switching - WTR)
MS O
Clear N/A
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
G (Switching - MS) LP C (Switching - LP)
LW O - if on another link
D (Idle - LW) - if on the same
link
FS E (Switching - FS)
SF F (Switching - SF)
Recover from SF N/A
MS N/A - if on the same link
G (Switching - MS) - if on
another link release the
switches but signal MS
Clear A
WTR expires N/A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
H (Switching - WTR) LP C (Switching - LP)
LW D (Idle - W)
FS E (Switching - FS)
SF F (Switching - SF)
Recover from SF N/A
MS G (Switching - MS)
Clear A
WTR expires A
EXER O
=====================================================================
Initial state New request New state
------------- ----------- ---------
I (Switching - EXER) LP C (Switching - LP)
LW D (idle - W)
FS E (Switching - FS)
SF F (Switching - SF)
Recover from SF N/A
Cheng, et al. Expires December 14, 2017 [Page 41]
Internet-Draft MSRP June 2017
MS G (Switching - MS)
Clear A
WTR expires N/A
EXER N/A - if on the same link
I (Switching - EXER)
=====================================================================
5.3.4. State Transitions When Remote Request is Applied
The priority of a remote request does not depend on the side from
which the request is received.
=====================================================================
Initial state New request New state
------------- ----------- ---------
A (Idle) LP C (Switching - LP)
FS E (Switching - FS)
SF F (Switching - SF)
MS G (Switching - MS)
WTR N/A
EXER I (Switching - EXER)
RR N/A
NR A (Idle)
=====================================================================
Initial state New request New state
------------- ----------- ---------
B (Pass-through) LP C (Switching - LP)
FS N/A - cannot happen when there
is LP request in the ring
E (Switching - FS) - otherwise
SF N/A - cannot happen when there
is LP request in the ring
F (Switching - SF) - otherwise
MS N/A - cannot happen when there
is LP, FS or SF request
in the ring
G (Switching - MS) - otherwise
WTR N/A - cannot happen when there
is LP, FS, SF or MS
request in the ring
EXER N/A - cannot happen when there
is LP, FS, SF, MS or WTR
request in the ring
I (Switching - EXER) -
otherwise
RR N/A
NR A (Idle) - if received from
Cheng, et al. Expires December 14, 2017 [Page 42]
Internet-Draft MSRP June 2017
both sides
=====================================================================
Initial state New request New state
------------- ----------- ---------
C (Switching - LP) LP C (Switching - LP)
FS N/A - cannot happen when there
is LP request in the ring
SF N/A - cannot happen when there
is LP request in the ring
MS N/A - cannot happen when there
is LP request in the ring
WTR N/A
EXER N/A - cannot happen when there
is LP request in the ring
RR C (Switching - LP)
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
D (Idle - LW) LP C (Switching - LP)
FS E (Switching - FS)
SF F (Switching - SF)
MS G (Switching - MS)
WTR N/A
EXER I (Switching - EXER)
RR N/A
NR D (Idle - LW)
=====================================================================
Initial state New request New state
------------- ----------- ---------
E (Switching - FS) LP C (Switching - LP)
FS E (Switching - FS)
SF E (Switching - FS)
MS N/A - cannot happen when there
is FS request in the ring
WTR N/A
EXER N/A - cannot happen when there
is FS request in the ring
RR E (Switching - FS)
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
F (Switching - SF) LP C (Switching - LP)
FS F (Switching - SF)
SF F (Switching - SF)
MS N/A - cannot happen when there
is SF request in the ring
Cheng, et al. Expires December 14, 2017 [Page 43]
Internet-Draft MSRP June 2017
WTR N/A
EXER N/A - cannot happen when there
is SF request in the ring
RR F (Switching - SF)
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
G (Switching - MS) LP C (Switching - LP)
FS E (Switching - FS)
SF F (Switching - SF)
MS G (Switching - MS) - release
the switches but signal MS
WTR N/A
EXER N/A - cannot happen when there
is MS request in the ring
RR G (Switching - MS)
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
H (Switching - WTR) LP C (Switching - LP)
FS E (Switching - FS)
SF F (Switching - SF)
MS G (Switching - MS)
WTR H (Switching - WTR)
EXER N/A - cannot happen when there
is WTR request in the ring
RR H (Switching - WTR)
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
I (Switching - EXER) LP C (Switching - LP)
FS E (Switching - FS)
SF F (Switching - SF)
MS G (Switching - MS)
WTR N/A
EXER I (Switching - EXER)
RR I (Switching - EXER)
NR N/A
=====================================================================
Cheng, et al. Expires December 14, 2017 [Page 44]
Internet-Draft MSRP June 2017
5.3.5. State Transitions When Request Addresses to Another Node is
Received
The priority of a remote request does not depend on the side from
which the request is received.
=====================================================================
Initial state New request New state
------------- ----------- ---------
A (Idle) LP B (Pass-through)
FS B (Pass-through)
SF B (Pass-through)
MS B (Pass-through)
WTR B (Pass-through)
EXER B (Pass-through)
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
B (Pass-through) LP B (Pass-through)
FS N/A - cannot happen when there
is LP request in the ring
B (Pass-through) - otherwise
SF N/A - cannot happen when there
is LP request in the ring
B (Pass-through) - otherwise
MS N/A - cannot happen when there
is LP, FS or SF request
in the ring
B (Pass-through) - otherwise
WTR N/A - cannot happen when there
is LP, FS, SF or MS
request in the ring
B (Pass-through) - otherwise
EXER N/A - cannot happen when there
is LP, FS, SF, MS or WTR
request in the ring
B (Pass-through) - otherwise
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
C (Switching - LP) LP C (Switching - LP)
FS N/A - cannot happen when there
is LP request in the ring
SF N/A - cannot happen when there
Cheng, et al. Expires December 14, 2017 [Page 45]
Internet-Draft MSRP June 2017
is LP request in the ring
MS N/A - cannot happen when there
is LP request in the ring
WTR N/A - cannot happen when there
is LP in the ring
EXER N/A - cannot happen when there
is LP request in the ring
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
D (Idle - LW) LP B (Pass-through)
FS B (Pass-through)
SF B (Pass-through)
MS B (Pass-through)
WTR B (Pass-through)
EXER B (Pass-through)
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
E (Switching - FS) LP B (Pass-through)
FS E (Switching - FS)
SF E (Switching - FS)
MS N/A - cannot happen when there
is FS request in the ring
WTR N/A - cannot happen when there
is FS request in the ring
EXER N/A - cannot happen when there
is FS request in the ring
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
F (Switching - SF) LP B (Pass-through)
FS F (Switching - SF)
SF F (Switching - SF)
MS N/A - cannot happen when there
is SF request in the ring
WTR N/A - cannot happen when there
is SF request in the ring
EXER N/A - cannot happen when there
is SF request in the ring
RR N/A
NR N/A
Cheng, et al. Expires December 14, 2017 [Page 46]
Internet-Draft MSRP June 2017
=====================================================================
Initial state New request New state
------------- ----------- ---------
G (Switching - MS) LP B (Pass-through)
FS B (Pass-through)
SF B (Pass-through)
MS G (Switching - MS) - release
the switches but signal MS
WTR N/A - cannot happen when there
is MS request in the ring
EXER N/A - cannot happen when there
is MS request in the ring
RR N/A
NR N/A
=====================================================================
Initial state New request New state
------------- ----------- ---------
H (Switching - WTR) LP B (Pass-through)
FS B (Pass-through)
SF B (Pass-through)
MS B (Pass-through)
WTR N/A
EXER N/A - cannot happen when there
is WTR request in the ring
RR N/A
NR N/A
=====================================================================
Initial state New request New state
I (Switching - EXER) LP B (Pass-through)
FS B (Pass-through)
SF B (Pass-through)
MS B (Pass-through)
WTR N/A
EXER I (Switching - EXER)
RR N/A
NR N/A
=====================================================================
6. IANA Considerations
IANA is requested to administer the assignment of new values defined
in this document and listed in the sections below.
Cheng, et al. Expires December 14, 2017 [Page 47]
Internet-Draft MSRP June 2017
6.1. G-ACh Channel Type
The Channel Types for the Generic Associated channel (GACh) are
allocated from the IANA PW Associated Channel Type registry defined
in [RFC4446] and updated by [RFC5586].
IANA is requested to allocate a new GACh Channel Type as follows:
Value| Description | Reference
------+---------------------------+--------------
TBD | Ring Protection Switching |this document
| Protocol (RPS) |
------+---------------------------+--------------
6.2. RPS Request Codes
IANA is requested to create a new sub-registry under the
"Multiprotocol Label Switching (MPLS) Operations, Administration, and
Management (OAM) Parameters" registry called the "MPLS RPS Request
Code Registry". All code points within this registry shall be
allocated according to the "Specification Required" procedure as
specified in [RFC5226].
The RPS Request Field is 8 bits, the allocated values are as follows:
Value Description Reference
------- --------------------------- ---------------
0 No Request (NR) this document
1 Reverse Request (RR) this document
2 unassigned
3 Exercise (EXER) this document
4 unassigned
5 Wait-To-Restore (WTR) this document
6 Manual Switch (MS) this document
7-10 unassigned
11 Signal Fail (SF) this document
12 unassigned
13 Forced Switch (FS) this document
14 unassigned
15 Lockout of Protection (LP) this document
16-254 unassigned
255 Reserved
7. Operational Considerations
This document describes three protection modes of the RPS protocol.
Operators could choose the appropriate protection mode according to
their network and service requirement.
Cheng, et al. Expires December 14, 2017 [Page 48]
Internet-Draft MSRP June 2017
Wrapping mode provides a ring protection mechanism in which the
protected traffic will reach every node of the ring, and is
applicable to protect both the point-to-point LSPs and LSPs which
needs to be dropped in several ring nodes, i.e. the point-to-
multipoint applications. When protection is in active, the protected
traffic is switched (wrapped) to/from the protection ring tunnel at
both sides of the defective link/node. Due to the wrapping the
additional propagation delay and bandwidth consumption of the
protection tunnel are considerable. For bidirectional LSP, the
protected traffic in both directions is co-routed.
Short wrapping mode provides a ring protection mechanism which can be
used to protect only point-to-point LSPs. When protection is in
active, the protected traffic wrapped to the protection ring tunnel
at the defective link/node and leaves the ring when the protection
ring tunnel reach the egress node. Compared with wrapping mode,
short wrapping can reduce the propagation latency and bandwidth
consumption of the protection tunnel. However the two directions of
a protected bidirectional LSP are not totally co-routed.
Steering mode provides a ring protection mechanism that can be used
to protect only point-to-point LSPs. When protection is in
active,the protected traffic is switched to the protection ring
tunnel at the ingress node and leaves the ring when the protection
ring tunnel reach the egress node. Steering mode has the least
propagation delay and bandwidth consumption of the three modes, and
the two directions of a protected bidirectional LSP can be kept co-
routed.
Note that only one protection mode can be provisioned in the whole
ring for all protected traffic.
8. Security Considerations
MPLS-TP is a subset of MPLS and so builds upon many of the aspects of
the security model of MPLS. Please refer to [RFC5920] for generic
MPLS security issues and methods for securing traffic privacy and
integrity.
The RPS message defined in this document is used for protection
coordination on the ring, if it is injected or modified by an
attacker, the ring nodes might not agree on the protection action,
and the improper protection switching action may cause temporary
break to services traversing the ring. It is important that the RPS
message is used within a trusted MPLS-TP network domain as described
in [RFC6941].
Cheng, et al. Expires December 14, 2017 [Page 49]
Internet-Draft MSRP June 2017
The RPS message is carried in the G-ACh [RFC5586], so it is dependent
on the security of the G-ACh itself. The G-ACh is a generalization
of the Associated Channel defined in [RFC4385]. Thus, this document
relies on the security mechanisms provided for the Associated Channel
as described in those two documents.
As described in the security considerations of [RFC6378], the G-ACh
is essentially connection oriented so injection or modification of
control messages requires the subversion of a transit node. Such
subversion is generally considered hard in connection oriented MPLS
networks and impossible to protect against at the protocol level.
Management level techniques are more appropriate. The procedures and
protocol extensions defined in this document do not affect the
security model of MPLS-TP linear protection as defined in [RFC6378].
9. Contributing Authors
Cheng, et al. Expires December 14, 2017 [Page 50]
Internet-Draft MSRP June 2017
Kai Liu
Huawei Technologies
Email: alex.liukai@huawei.com
Jia He
Huawei Technologies
Email: hejia@huawei.com
Fang Li
China Academy of Telecommunication Research MIIT., China
Email: lifang@catr.cn
Jian Yang
ZTE Corporation P.R.China
Email: yang.jian90@zte.com.cn
Junfang Wang
Fiberhome Telecommunication Technologies Co., LTD.
Email: wjf@fiberhome.com.cn
Wen Ye
China Mobile
Email: yewen@chinamobile.com
Minxue Wang
China Mobile
Email: wangminxue@chinamobile.com
Sheng Liu
China Mobile
Email: liusheng@chinamobile.com
Guanghui Sun
Huawei Technologies
Email: sunguanghui@huawei.com
10. Acknowledgements
The authors would like to thank Gregory Mirsky, Yimin Shen, Eric
Osborne, Spencer Jackson and Eric Gray for their valuable comments
and suggestions.
11. References
Cheng, et al. Expires December 14, 2017 [Page 51]
Internet-Draft MSRP June 2017
11.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>.
[RFC3031] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol
Label Switching Architecture", RFC 3031,
DOI 10.17487/RFC3031, January 2001,
<http://www.rfc-editor.org/info/rfc3031>.
[RFC4385] Bryant, S., Swallow, G., Martini, L., and D. McPherson,
"Pseudowire Emulation Edge-to-Edge (PWE3) Control Word for
Use over an MPLS PSN", RFC 4385, DOI 10.17487/RFC4385,
February 2006, <http://www.rfc-editor.org/info/rfc4385>.
[RFC4446] Martini, L., "IANA Allocations for Pseudowire Edge to Edge
Emulation (PWE3)", BCP 116, RFC 4446,
DOI 10.17487/RFC4446, April 2006,
<http://www.rfc-editor.org/info/rfc4446>.
[RFC5586] Bocci, M., Ed., Vigoureux, M., Ed., and S. Bryant, Ed.,
"MPLS Generic Associated Channel", RFC 5586,
DOI 10.17487/RFC5586, June 2009,
<http://www.rfc-editor.org/info/rfc5586>.
[RFC5654] Niven-Jenkins, B., Ed., Brungard, D., Ed., Betts, M., Ed.,
Sprecher, N., and S. Ueno, "Requirements of an MPLS
Transport Profile", RFC 5654, DOI 10.17487/RFC5654,
September 2009, <http://www.rfc-editor.org/info/rfc5654>.
11.2. Informative References
[RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an
IANA Considerations Section in RFCs", BCP 26, RFC 5226,
DOI 10.17487/RFC5226, May 2008,
<http://www.rfc-editor.org/info/rfc5226>.
[RFC5920] Fang, L., Ed., "Security Framework for MPLS and GMPLS
Networks", RFC 5920, DOI 10.17487/RFC5920, July 2010,
<http://www.rfc-editor.org/info/rfc5920>.
[RFC6371] Busi, I., Ed. and D. Allan, Ed., "Operations,
Administration, and Maintenance Framework for MPLS-Based
Transport Networks", RFC 6371, DOI 10.17487/RFC6371,
September 2011, <http://www.rfc-editor.org/info/rfc6371>.
Cheng, et al. Expires December 14, 2017 [Page 52]
Internet-Draft MSRP June 2017
[RFC6378] Weingarten, Y., Ed., Bryant, S., Osborne, E., Sprecher,
N., and A. Fulignoli, Ed., "MPLS Transport Profile (MPLS-
TP) Linear Protection", RFC 6378, DOI 10.17487/RFC6378,
October 2011, <http://www.rfc-editor.org/info/rfc6378>.
[RFC6941] Fang, L., Ed., Niven-Jenkins, B., Ed., Mansfield, S., Ed.,
and R. Graveman, Ed., "MPLS Transport Profile (MPLS-TP)
Security Framework", RFC 6941, DOI 10.17487/RFC6941, April
2013, <http://www.rfc-editor.org/info/rfc6941>.
[RFC6974] Weingarten, Y., Bryant, S., Ceccarelli, D., Caviglia, D.,
Fondelli, F., Corsi, M., Wu, B., and X. Dai,
"Applicability of MPLS Transport Profile for Ring
Topologies", RFC 6974, DOI 10.17487/RFC6974, July 2013,
<http://www.rfc-editor.org/info/rfc6974>.
Authors' Addresses
Weiqiang Cheng
China Mobile
Email: chengweiqiang@chinamobile.com
Lei Wang
China Mobile
Email: wangleiyj@chinamobile.com
Han Li
China Mobile
Email: lihan@chinamobile.com
Huub van Helvoort
Hai Gaoming BV
Email: huubatwork@gmail.com
Jie Dong
Huawei Technologies
Email: jie.dong@huawei.com
Cheng, et al. Expires December 14, 2017 [Page 53]