RSTP Convergence

The convergence of STP in a network is the process that takes all switches from a state of independence (each thinks it must be the STP root) to one of uniformity, in which each switch has a place in a loop-free tree topology. You can think of convergence as a two-stage process:

1. One common Root Bridge must be "elected," and all switches must know about it.

2. The state of every switch port in the STP domain must be brought from a Blocking state to the appropriate state to prevent loops.

Convergence generally takes time because messages are propagated from switch to switch. The traditional 802.1D STP also requires the expiration of several timers before switch ports can safely be allowed to forward data.

RSTP takes a different approach when a switch needs to decide how to participate in the tree topology. When a switch first joins the topology (perhaps it was just powered up) or has detected a failure in the existing topology, RSTP requires it to base its forwarding decisions on the type of port.

Port Types

Every switch port can be considered one of the following types:

■ Edge port—A port at the "edge" of the network, where only a single host connects. Traditionally, this has been identified by enabling the STP PortFast feature. RSTP keeps the PortFast concept for familiarity. By definition, the port cannot form a loop as it connects to one host, so it can be placed immediately in the Forwarding state. However, if a BPDU ever is received on an edge port, the port immediately loses its edge port status.

■ Root port—The port that has the best cost to the root of the STP instance. Only one root port can be selected and active at any time, although alternative paths to the root can exist through other ports. If alternative paths are detected, those ports are identified as alternate root ports and immediately can be placed in the Forwarding state when the existing root port fails.

■ Point-to-point port—Any port that connects to another switch and becomes a designated port. A quick handshake with the neighboring switch, rather than a timer expiration, decides the port state. BPDUs are exchanged back and forth in the form of a proposal and an agreement. One switch proposes that its port becomes a designated port; if the other switch agrees, it replies with an agreement message.

Point-to-point ports automatically are determined by the duplex mode in use. Full-duplex ports are considered point to point because only two switches can be present on the link. STP convergence can occur quickly over a point-to-point link through RSTP handshake messages.

Half-duplex ports, on the other hand, are considered to be on a shared medium with possibly more than two switches present. They are not point-to-point ports. STP convergence on a half-duplex port must occur between several directly connected switches. Therefore, the traditional 802.1D style convergence must be used. This results in a slower response because the shared-medium ports must go through the fixed Listening and Learning state time periods.

It's easy to see how two switches quickly can converge to a common idea of which one is the root and which one will have the designated port after just a single exchange of BPDUs. What about a larger network, where 802.1D BPDUs normally would have to be relayed from switch to switch?

RSTP handles the complete STP convergence of the network as a propagation of handshakes over point-to-point links. When a switch needs to make an STP decision, a handshake is made with the nearest neighbor. When that is successful, the handshake sequence is moved to the next switch and the next, as an ever-expanding wave moving toward the network's edges.

During each handshake sequence, a switch must take measures to completely ensure that it will not introduce a bridging loop before moving the handshake outward. This is done through a synchronization process.


To participate in RSTP convergence, a switch must decide the state of each of its ports. Nonedge ports begin in the Discarding state. After BPDUs are exchanged between the switch and its neighbor, the Root Bridge can be identified. If a port receives a superior BPDU from a neighbor, that port becomes the root port.

For each nonedge port, the switch exchanges a proposal-agreement handshake to decide the state of each end of the link. Each switch assumes that its port should become the designated port for the segment, and a proposal message (a configuration BPDU) is sent to the neighbor suggesting this.

When a switch receives a proposal message on a port, the following sequence of events occurs Figure 11-1 shows the sequence, based on the center Catalyst switch:

1. If the proposal's sender has a superior BPDU, the local switch realizes that the sender should be the designated switch (having the designated port) and that its own port must become the new root port.

2. Before the switch agrees to anything, it must synchronize itself with the topology.

3. All nonedge ports immediately are moved into the Discarding (blocking) state so that no bridging loops can form.

4. An agreement message (a configuration BPDU) is sent back to the sender, indicating that the switch is in agreement with the new designated port choice. This also tells the sender that the switch is in the process of synchronizing itself.

5. The root port immediately is moved to the Forwarding state. The sender's port also immediately can begin forwarding.

6. For each nonedge port that is currently in the Discarding state, a proposal message is sent to the respective neighbor.

7. An agreement message is expected and received from a neighbor on a nonedge port.

8. The nonedge port immediately is moved to the Forwarding state.

Figure 11-1 Sequence of Events During RSTP Convergence

Figure 11-1 Sequence of Events During RSTP Convergence

Notice that the RSTP convergence begins with a switch sending a proposal message. The recipient of the proposal must synchronize itself by effectively isolating itself from the rest of the topology. All nonedge ports are blocked until a proposal message can be sent, causing the nearest neighbors to synchronize themselves. This creates a moving "wave" of synchronizing switches, which quickly can decide to start forwarding on their links only if their neighbors agree. Figure 11-2 shows how the synchronization wave travels through a network at three successive time intervals. Isolating the switches along the traveling wave inherently prevents bridging loops.

The entire convergence process happens quickly, at the speed of BPDU transmission, without the use of any timers. However, a designated port that sends a proposal message might not receive an agreement message reply. Suppose that the neighboring switch does not understand RSTP or has a problem replying. The sending switch then must become overly cautious and must begin playing by the 802.1D rules—the port must be moved through the legacy Listening and Learning states (using the Forward Delay timer) before moving to the Forwarding state.

Figure 11-2 RSTP Synchronization Traveling Through a Network

Figure 11-2 RSTP Synchronization Traveling Through a Network

+1 -1

Post a comment