Slide 1

Inter-network Ethernet Service Protection

Zehavit Alon Nurit SprecherJohn Lemon

Slide 2

Agenda

• Inter-network Ethernet Service Protection

–Overview

–Requirements

–Network architecture

▪ Possible connectivity constructions between Ethernet Networks

▪ Recommended construction

–Proposed solution

• Open discussion and next steps

Slide 3

Ethernet Services over Interconnected Networks

• Carrier Ethernet services are delivered over interconnected Ethernet networks - untagged, C-VLAN, S-VLAN, B-VLAN

• Interconnected networks may, for example, consist of:– a customer’s network connected to a service provider's network

– that is also connected to other service providers' networks.

• An end-to-end carrier Ethernet service can span several interconnected packet networks.

PBPBB-TE PBB PBB-TE PB

Slide 4

• Each Ethernet network may deploy a different packet transport technology which provides its own mechanisms aimed at ensuring network survivability. Examples are:

– Bridged Ethernet with MSTP or SPB or G.8032

– Traffic Engineered Ethernet with PBB-TE protection switching

• A protection mechanism is required for the interconnected zone.

PBPBB-TE PBB PBB-TE PB

Ethernet Services over Interconnected Networks

PB xSTP 1:1 SPB 1:1 PB xSTP

Interconnected Zone Interconnected Zone Interconnected Zone Interconnected Zone

Slide 5

Interconnected Networks Protection Mechanism: Requirements• Protect against any single failure or degradation of a facility

(link or node) in the interconnected zone

• Support all standard Ethernet frames: 802.1D, 802.1Q, 802.1ad, 802.1ah

• Support interconnection between different network types (e.g. CN-PBN, PBN-PBN, PBN-PBBN, PBBN-PBBN, etc.)

• Provide 50ms protection switching

• Provide a clear indication of the protection state

• Maintain an agnostic approach towards:– the Ethernet technology running on each of the interconnected

networks, and

– the protection mechanism deployed by each of the interconnected networks

Slide 6

Interconnected Networks Protection Mechanism: Requirements (cont’d.)

• Avoid modification of the protocols running inside each of the interconnected networks

• Ensure that multicast and broadcast frames are delivered only once over the interconnected zone

• Allow load balancing between the interfaces that connect the networks to ensure efficient utilization of resources

Slide 7

Possible Topologies

Mesh Ring

Slide 8

Dual Attached Connectivity

Three links are requiredTwo links are required

RingMesh

Slide 9

Enhanced Resiliency

Resiliency is enhanced by adding a node and two links, and by removing the redundant link. This operation may cause traffic disruption (if a facility fails during the upgrade operation).

Resiliency is enhanced by adding a node with dual attachment to the adjacent network. This provides protection against node failure (with no traffic disruption).

RingMesh

Dual attachment is widely deployed.

Slide 10

Connectivity between adjacent networks

Adjacent networks are connected by 8 connections:

2 direct connections A-D, B-C2 indirect connections A-D, B-C2 indirect connections B-D2 indirect connections A-C

The network local link may also be used to transmit internal traffic in the network (which may result in the utilization of BW required for protection).

Adjacent networks are connected by 4 direct (single-hop) connections: A-D, A-C, B-D, B-C

RingMesh

A

B C

D

A

B

D

C

Slide 11

Protection Path Load

Load sharing is supported across two links.

When a link connecting the networks fails, all traffic between the networks is transmitted via the other single link connecting the networks.

When a node fails, all traffic between the networks is transmitted via the other single link connecting the networks.

Load sharing is supported across all four links.

When a link fails, traffic is shared between the three other links.

When a node fails, traffic is shared between two links.

RingMesh

Slide 12

Load Sharing

Capable of supporting only two nodes in each network

Although nested rings are possible, they can significantly complicate the solution and the operation.

Capable of supporting more than two nodes and two links in each network, for connecting the networks with support for load sharing

RingMesh

Slide 13

Protection Path Cost

The cost of the protection path (in terms of the number of hops) is higher than that of the working path. (Revertive functionality is recommended.)

The cost of the protection path (in terms of the number of hops) is identical to that of the working path. (Revertive functionality is optional.)

RingMesh

Working

Protection

Slide 14

Multiple Failures

Mesh topology provides better resiliency in the event of multiple failures. Examples are:

RingMesh

Slide 15

Interconnection with Rings (G.8032)

A super loop is created.

Protection in the interconnection zone is not agnostic with regard to failures.

A mechanism is required to prevent the transmission of internal traffic from the network in the west (shown above) to the two nodes in the network in the east.

Protection in the interconnection zone is agnostic with regard to failures inside the ring.

RingMeshShared Link

G.8032 G.8032

Slide 16

Proposed Topologies

Mesh that supports dual-homing and that provides enhanced protection in the double dual-homing configuration

Slide 17

Interconnect zone

Solution Principles

4

31

2

7

A

6

5

C

B

D

Blue traffic (VLAN X) is only sent through port 1 (which is protected by port 2).

Blue traffic is sent through port 2 in the event of failure of link 1-3, or of node B

Blue traffic is sent through node C in the event that node A fails.

8

● The protection mechanism is available per Ethernet service in the interconnected zone (i.e. per VLAN).

● An Ethernet service is carried only over one of the interfaces which connects the two adjacent networks.

● In the event of a fault condition on the link or the peer node, traffic is redirected to the redundant interface.

● The service may also be protected by another node to avoid a single point of failure. If a node is no longer able to carry traffic, traffic is redirected over the redundant node.

Slide 18

Interconnect Area

Solution Principles

• The interconnected zone may include additional nodes, interfaces and links

• Each protected VLAN is configured, (independently of other VLANs) on: – Total of three nodes and four ports - on one of the networks, one node with two ports; on

the other network, two nodes with one port on each (i.e. dual-homing)

– Total of four nodes and eight ports - on both networks, two nodes with two ports each (i.e double dual-homing)

• Each protected VLAN can be transmitted over one out of two/four links. However, at any given time, it is only transmitted over one of the links crossing the interconnected zone.

4

31

2

8

7

A

6

5

C

B

D

E

10

11F11

9

1312

Slide 19

Solution Principles • For each protected VLAN, one of the nodes is responsible for selecting

the interface over which the traffic will be transmitted. This node functions as a master.

• The master is connected to two nodes. These two nodes follow the master’s decisions and function as slaves.

• The master node can be protected by a redundant node. In the event that the master fails, the redundant node functions as the master. This node is called a deputy. The deputy is connected to the same two slaves as the master.

M S

S

D

M

D

M

S

S

S

S

The role of each node (master, deputy and slave) is set for each VLAN by administrative configuration.

The same node may function as a master node for some VLANs (blue), as a deputy node for other VLANs (red), and a slave for other VLANs (green), thus enabling load sharing between the nodes.

D

Slide 20

Solution Principles

For each VLAN, the master/deputy/slave nodes are configured according to the following options:

• Additional parameters must be configured for the master and deputy nodes (not for the slaves):

– working port – the default port to use for traffic– protection port – the port to use when the working port can not be used.

M S

S

M

D

S

S

(a) (b)

S M

D

S

S

M

D

(c) (d)

Slide 21

Solution Principles

• The interface selection algorithm for each VLAN is based on – local configuration

– Information provided by link-level CCMs

• The protection state of all the protected VLANs is synchronized between peers by means of a single link-level CCM message.

12M

56D

78 S2

34 S1

Slave1 is active, and uses another port for VLAN X.

Master chooses the configured working port 1 for

VLAN X

Master is working so deputy does not

need to take over

Slave1 follows master’s decision and uses port 3

for VLAN X

Master uses this port for VLAN X

Master uses another port for VLAN X

Slave1 uses this port for VLAN X

Slave2 follows master’s decision and does not use any of its ports for VLAN X

Slave2 is not active for VLAN X

Deputy is not active for VLAN X

Deputy is not active for VLAN XSlave2 is not active for VLAN X

Slide 22

Solution Principles

• If a link fails, the master node uses the protection port (port 2) for VLAN X

12M

56D

78 S2

34 S1Slave is not active for

VLAN x

Slave2 is actctive and uses another port for VLAN X

Master uses this port for VLAN X

Deputy is npot active for VLAN X

Slave2 uses this port for VLAN XDeputy is not active for VLAN X

Slave1 does not receive anything from the master. It

does not use any of its ports for VLAN X

Master is working so deputy does not

need to take over

Slav2 follows master’s decision and uses port

7 for VLAN X

Link on port 1 is not working, Master chooses the configured protection

port 2 for VLAN X

Slide 23

Solution Principles

• If the master fails, the deputy is informed about it by the slaves and it becomes active

12M

56D

78 S2

34 S1

Slave1 does not work for VLAN X

Slave2 does not work for VLAN X

Deputy uses this port for VLAN X

Master failed. Does not send

anything

Does not receive anything from master so it doesn't use any port for VLAN X

Deputy sees that both slaved are not

working. It understands that the

master is not working so deputy takes over using its

working port (6)

Does not receive anything from master so it doesn't use any

port for VLAN

Slave2 follows deputy’s decision and uses port 8 for VLAN

X

Slave2 uses this this port for VLAN X

Deputy uses another port for VLAN X

does not use any of its ports for VLAN X

Slave1 does not work for VLAN X

Slide 24

Solution Principles

• A protected VLAN x is defined on 2 ports: On port A, VLAN x is configured as working entity, while on port B, VLAN x is configured as protection entity

• In a live system, the VLAN is transmitted only on one of the ports (working or protection entity).

• The 2 ports on which the VLAN is protected are grouped into a VLAN Protection Group (VPG). The VPG is a logical bridge port (as defined in 802.1Q + ad + ah).

Port A

VPG

VLAN x VLAN xPort B

Port AVLAN x VLAN x

Port B

Port AVLAN x Working VLAN x Protection

Port B

Slide 25

Solution Principles

• The VPG forwards VLAN traffic to the port selected by the algorithm.

• VLAN traffic received on a port is forwarded to the VPG. Learning occurs at the VPG level.

• The CCMs are sent and received by ports A and B, and the selection algorithm is implemented on the VPG, based on the information received on both ports.

VPG

Port AVLAN x VLAN x

Port B

VPG

Port AVLAN x VLAN x

Port B

Slide 26

Solution Principles

Location of the new shim

Slide 27

Intention

• Start a new project in the IEEE802.1 aimed at defining a protection mechanism for interconnected networks in the proposed topologies. The mechanism should comply with the requirements introduced in this presentation.

• Decide whether we should send a liaison to the MEF in order to receive feedback on (1) the proposed connectivity construction and (2) the requirements.

Slide 28

Thank You

zehavit.alon@nsn.comnurit.sprecher@nsn.comjlemon@ieee.org