A REP segment is a chain of ports connected to each other and configured with a segment ID. Each segment consists of standard (nonedge) segment ports and two user-configured edge ports. A router can have no more than two ports that belong to the same segment, and each segment port can have only one external neighbor. A segment can go through a shared medium, but on any link, only two ports can belong to the same segment. REP is supported only on Trunk Ethernet Flow Point (EFP) interfaces.

The figure below shows an example of a segment consisting of six ports spread across four switches. Ports E1 and E2 are configured as edge ports. When all ports are operational (as in the segment on the left), a single port is blocked, shown by the diagonal line. When there is a failure in the network, the blocked port returns to the forwarding state to minimize network disruption.

Figure 1. REP Open Segments

The segment shown in the figure above is an open segment; there is no connectivity between the two edge ports. The REP segment cannot cause a bridging loop, and you can safely connect the segment edges to any network. All hosts connected to routers inside the segment have two possible connections to the rest of the network through the edge ports, but only one connection is accessible at any time. If a failure occurs on any segment or on any port on a REP segment, REP unblocks all ports to ensure that connectivity is available through the other gateway.

The segment shown in the figure below is a ring segment, and it has both edge ports located on the same router. With this configuration, you can create a redundant connection between any two routers in the segment.

Figure 2. REP Ring Segment

REP segments have the following characteristics:

• If all ports in a segment are operational, one port (referred to as the alternate port) is in the blocked state for each VLAN. If VLAN load balancing is configured, two ports in the segment control the blocked state of VLANs.

• If one or more ports in a segment is not operational, and cause a link failure, all ports forward traffic on all VLANs to ensure connectivity.

• In case of a link failure, alternate ports are unblocked as quickly as possible. When the failed link is up, a logically blocked port per VLAN is selected with minimal disruption to the network.

You can construct almost any type of network based on REP segments. REP also supports VLAN load balancing, which is controlled by the primary edge port but can occurring at any port in the segment.

REP does not use an end-to-end polling mechanism between edge ports to verify link integrity. It implements local link failure detection. When enabled on an interface, the REP Link Status Layer (LSL) detects its REP-aware neighbor and establishes connectivity within the segment. All VLANs are blocked on an interface until the REP LSL detects the neighbor. After the neighbor is identified, REP determines which neighbor port should become the alternate port and which ports should forward traffic.

Each port in a segment has a unique port ID. The port ID format is similar to that used by the spanning tree algorithm: a port number (unique on the bridge), associated to a MAC address (unique in the network). When a segment port is up, LSL sends packets that include the segment ID and the port ID. The port is declared as operational after it performs a three-way handshake with a neighbor in the same segment. A segment port does not become operational under the following conditions:

• No neighbor has the same segment ID.

• More than one neighbor has the same segment ID.

• The neighbor does not acknowledge the local port as a peer.

Each port creates an adjacency with its immediate neighbor. Once the neighbor adjacencies are created, the ports negotiate to determine one blocked port for the segment, which is the alternate port. All other ports become unblocked. By default, REP packets are sent to a PortFast Bridge Protocol Data Unit (BPDU) class MAC address. The packets can also be sent to the Cisco multicast address, which at present is used only to send blocked port advertisement (BPA) messages when there is a failure in the segment. The packets are dropped by devices not running REP.

Because REP runs on a physical-link basis and not on a per-VLAN basis, only one hello message is required for all VLANs, thus reducing the load on the protocol. We recommend that you create VLANs consistently on all switches in a given segment and configure VLANs on REP trunk ports. To avoid the delay introduced by relaying messages in software, REP also allows some packets to be flooded to a regular multicast address. These messages operate at the hardware flood layer (HFL) and are flooded to the whole network, not just the REP segment. Switches that do not belong to the segment treat the messages as data traffic. You can control flooding of these messages by configuring a dedicated administrative VLAN for the whole domain.

The estimated convergence recovery time is less than 200 milliseconds (ms) for the local segment.

One edge port in a REP segment acts as the primary edge port and the other as the secondary edge port. It is the primary edge port that always participates in VLAN load balancing in the segment. REP VLAN load balancing is achieved by blocking some VLANs at a configured alternate port and all other VLANs at the primary edge port. When you configure VLAN load balancing, you can specify the alternate port using any one of the following ways:

• By entering the port ID of the interface. To identify the port ID of a port in the segment, enter the show interface rep detail command for the port.

• By entering the neighbor offset number of a port in the segment, which identifies the downstream neighbor port of an edge port. The neighbor offset number range is -256 to +256; a value of 0 is invalid. The primary edge port has an offset number of 1; positive numbers above 1 identify downstream neighbors of the primary edge port. Negative numbers indicate the secondary edge port (offset number -1) and its downstream neighbors.

  Note	You configure offset numbers on the primary edge port by identifying a port’s downstream position from the primary (or secondary) edge port. You cannot enter an offset value of 1 because 1 is the offset number of the primary edge port .

• By entering the preferred keyword to select the port that you previously configured as the preferred alternate port in the rep segment preferred command.

When the REP segment is complete, all VLANs are blocked. VLAN load balancing can be triggered in one of the following two ways:

• You can manually trigger VLAN load balancing at any time by entering the rep preempt segment segment-id command on the router that has the primary edge port.

• You can configure a preempt delay time by entering the rep preempt delay seconds command. After a link failure and recovery, VLAN load balancing begins after the configured preemption time period elapses. The delay timer restarts if another port fails before the time has elapsed.

  Note	A VLAN load balancing does not start working until triggered by either a manual intervention or a link failure and recovery.

When VLAN load balancing is triggered, the primary edge port sends out a message to alert all interfaces in the segment about the preemption. When the message is received by the secondary edge port, a message is generated in the network to notify the alternate port to block the set of VLANs specified in the message and to notify the primary edge port to block the remaining VLANs.

You can also configure a particular port in the segment to block all VLANs. VLAN load balancing is initiated only by the primary edge port and is not possible if the segment is not terminated by an edge port on each end. The primary edge port determines the local VLAN load balancing configuration.

To reconfigure VLAN load balancing, you must reconfigure the primary edge port. When you change the VLAN-load balancing configuration, the primary edge port again waits for the rep preempt segment command or for the configured preempt delay period after a port failure and recovery before executing the new VLAN load balancing configuration. If you change an edge port to a regular segment port, the existing VLAN load balancing status does not change. Configuring a new edge port might cause a new topology configuration.

REP does not interact with STP or with Flex Links but can coexist with both of them. A port that belongs to a segment is removed from spanning tree control, and STP BPDUs are not accepted or sent from segment ports. Therefore, STP cannot run on a segment.

To migrate from an STP ring configuration to a REP segment configuration, begin by configuring a single port in the ring as part of the segment and continue by configuring contiguous ports to minimize the number of segments. Each segment always contains a blocked port, so multiple segments mean multiple blocked ports and a potential loss of connectivity. You can configure the edge ports when the segment has been configured in both directions up to the location of the edge ports.

Ports in REP segments take one of following three roles or states: Failed, Open, or Alternate.

• A port configured as a regular segment port starts as a failed port.

• After neighbor adjacencies are determined, the port transitions to the alternate port state, blocking all VLANs on the interface. Blocked port negotiations occur, and when the segment settles, one blocked port remains in the alternate role, and all other ports become open ports.

• When a failure occurs in a link, all ports move to the failed state. When the alternate port receives the failure notification, the port changes to the open state forwarding all VLANs.

A regular segment port converted to an edge port, or an edge port converted to a regular segment port, does not always result in a topology change. If you convert an edge port into a regular segment port, VLAN load balancing is not implemented unless it has been configured. For VLAN load balancing, you must configure two edge ports in the segment.

A segment port that is reconfigured as a spanning tree port restarts according to the spanning tree configuration. By default, this port is a designated blocking port. If the PortFast BPDU Guard Enhancement feature is configured or if STP is disabled, the port goes into the forwarding state.

Normally, in a Virtual Private LAN Service (VPLS) network core, all nodes are connected in a full-mesh topology and each node has connectivity to all other nodes. In the full-mesh topology, there is no need for a node to retransmit data to another node. In Figure 3, the common ring provides a path where the packet can be forwarded to another network provider edge (N-PE) router, breaking split horizon model.

REP emulates a common link connection the REP ring supports the VPLS full-mesh model, but maintains the split horizon properties so the super-loop does not exist. The emulated common link uses the Clustering over the WAN (CWAN) line card, which is also used for the VPLS uplink. This emulated common link forwards data from the ring to either the VPLS uplink or to the other side of the ring; blocks data coming from the VPLS core network; and handles access to pseudowire for Hierarchical-VPLS (H-VPLS) topologies.

REP is disabled on all interfaces. When enabled, the interface is a regular segment port unless it is configured as an edge port.

When REP is enabled, the sending of segment topology change notices (STCNs) is disabled, all VLANs are blocked, and the administrative VLAN is VLAN 1.

When VLAN load balancing is enabled, the default is manual preemption with the delay timer disabled. If VLAN load balancing is not configured, the default after manual preemption is to block all VLANs at the primary edge port.

A segment is a collection of ports connected in a chain and configured with a segment ID. To configure REP segments, you should configure the REP administrative VLAN (or use the default VLAN 1) and then add ports to the segment in interface configuration mode. You should configure two edge ports in the segment, with one as the primary edge port and the other, by default, as the secondary edge port. A segment has only one primary edge port. If you configure two ports in a segment as primary edge ports, for example, ports on different switches, REP selects one of them to serve as the primary edge port. You can also optionally configure where to send segment STCNs and VLAN load balancing. For more information about configuring REP Administrative VLANs, see the Configuring the REP Administrative VLAN section.

Follow these guidelines when configuring REP:

• We recommend that you begin by configuring one port and then configure contiguous ports to minimize the number of segments and the number of blocked ports.

• If more than two ports in a segment fail when no external neighbors are configured, one port goes into a forwarding state for the data path to help maintain connectivity during configuration. In the show rep interface command output, the Port Role for this port shows as “Fail Logical Open”; the Port Role for the other failed port shows as “Fail No Ext Neighbor”. When the external neighbors for the failed ports are configured, the ports go through the alternate port state transitions and eventually go to an open state or remain as the alternate port, based on the alternate port selection mechanism.

• REP ports must be Layer 2 IEEE 802.1Q or Trunk EFP ports.

• We recommend that you configure all trunk ports in the segment with the same set of allowed VLANs.

• Be careful when configuring REP through a Telnet connection. Because REP blocks all VLANs until another REP interface sends a message to unblock it. You might lose connectivity to the router if you enable REP in a Telnet session that accesses the router through the same interface.

• You cannot run REP and STP on the same segment or interface.

• If you connect an STP network to a REP segment, be sure that the connection is at the segment edge. An STP connection that is not at the edge could cause a bridging loop because STP does not run on REP segments. All STP BPDUs are dropped at REP interfaces.

• If REP is enabled on two ports on a router, both ports must be either regular segment ports or edge ports. REP ports follow these rules:

  • If only one port on a router is configured in a segment, the port should be an edge port.
  • If two ports on a router belong to the same segment, both ports must be edge ports or must be regular segment ports.
  • If two ports on a router belong to the same segment and one is configured as an edge port and the other as a regular segment port (a misconfiguration), the edge port is treated as a regular segment port.

• REP interfaces come up in a blocked state and remain in a blocked state until they are safe to be unblocked. You need to be aware of this status to avoid sudden connection losses.

• REP ports cannot be configured as one of the following port types:

  • Switched Port Analyzer (SPAN) destination port
  • Private VLAN port
  • Tunnel port
  • Access port

• There can be a maximum of 22 REP segments per router.

Resilient Ethernet Protocol (REP) can be configured on Trunk EFP ports at the interface level on Cisco ASR 903 Series Routers. Trunk EFP ports can have several bridged VLAN services running on them. VLANs can be set to blocking and forwarding state on a Trunk EFP port. A user must enable REP on a port. By default, REP is disabled on all ports.

In a ring network topology, the Fast Last Link Status (LSL) process detects a neighboring port and maintains a connection with it. The timer on a port can be configured within 200-10000 ms to receive LSL frames. If no LSL frames are received from 200 to10000 ms from the neighboring port, the link between routers is considered as down. The tear-down operation and action is taken to bring up the link and restore traffic.

In the ring network topology, REP might fail to converge the traffic within 50 ms. For example, if the topology is made of copper cable, REP might fail to converge the traffic due to hardware limitations of the copper interface. In such a scenario, a remote end can take up to 700 ms to detect shutdown failure of a local port. The REP LSL is enhanced to achieve higher timer granularity and faster failure detection on the remote side.

The figure below shows the delay in failure detection due to hardware limitation of a Copper interface.

Figure 3. Delay in Failure Detection

When a router crashes, it takes between 3 to 5 seconds for the router to get into active mode and start sending REP Fast Hello packets. If the value of the age out timer configured by the lsl age out timer command is less than 3 seconds, the remote end detects a port failure and reconverges. After reconverging, the router sends out a BPDU with a special type, length, and, value (TLV) to the connected port. The router learns the port’s local and remote sequence number so that the subsequent REP three-way link integrity check does not fail. The Stateful Switchover (SSO) support for REP ensures that a Fast Hello packet can be sent from the router before the LSL interval expires.

In a ring network topology, aggregation nodes do not support REP. A REP segment can be created with no-neighbor ports to achieve convergence of switches. The figure below shows P1 and P2 as Edge No-Neighbor ports in a ring topology. In this configuration P1 and P2 can block traffic. If there is a failure on any of the links, all the switches with REP configuration converge. Since P1 and P2 are not edges, they do not support the following tasks:

• Perform VLAN load balancing.

• Detect topology changes to other segments and the Spanning Tree Protocol (STP).

• Choose the port that can preempt.

• Display the complete segment topology.

The Edge No-Neighbor support enables defining a new type of edge that has an internal neighbor. In the figure below, P1 and P2 are configured as Edge No-Neighbor ports rather than intermediate segment ports. These ports inherit properties of edge ports and overcome the limitations listed above. Thus, the Edge No-Neighbor port (P1 or P2) can send the Multiple Spanning Tree (MST) protocol, a Topology Change Notification (TCN), and a REP TCN for another segment towards the aggregation switch.

Figure 4. Ring Topology with Edge No-Neighbor Ports