The previous section mentioned that a
bridge spends 15 seconds in each of the listening and learning
states by default. In all, the Spanning-Tree Protocol is controlled
by the three timers shown in Figure .
The "hello time" controls
the time interval between the sending of configuration BPDUs. The
802.1D standard specifies a default value of 2 seconds. This value
really controls only configuration BPDUs as they are generated at
the root bridge-other bridges propagate BPDUs from the root bridge
as they are received. In other words, if BPDUs stop arriving for
2-20 seconds because of a network disturbance, non-root bridges stop
sending periodic BPDUs during this time. If the outage lasts for
more than 20 seconds, the default "max age" time, the
bridge invalidates the saved BPDUs and begins looking for a new root
port.
"Forward Delay" is the time
that the bridge spends in the listening and learning states. This is
a single value that controls both states. The default value of 15
seconds was originally derived assuming a maximum network size of
seven bridge hops, a maximum of three lost BPDUs, and a hello-time
interval of 2 seconds. The forward-delay timer also controls the
bridge table age-out period after a change in the active topology.
"Max age" is the time that
a bridge stores a BPDU before discarding it. Recall that each port
saves a copy of the best BPDU it has seen. As long as the bridge
receives a continuous stream of BPDUs every 2 seconds, the receiving
bridge maintains a continuous copy of the BPDU values. However, if
the device sending this best BPDU fails, some mechanism must exist
to allow other bridges to take over.
For example, assume that the Segment
3 link in Figure
uses a hub and the Cat-B Port-1/2 transceiver fails. Cat-C has no
immediate notification of the failure because it's still receiving
Ethernet "link" from the hub. The only thing Cat-C knows
is that BPDUs stop arriving. Twenty seconds (max age) after the
failure, Cat-C Port-1/2 ages out the BPDU information that lists
Cat-B as having the best designated port for Segment 3. This forces
Cat-C:Port-1/2 to transition into the listening state in an effort
to become the designated port. Because Cat-C Port-1/2 now offers the
most attractive access from the root bridge to this link, it
eventually transitions all the way into forwarding mode. In
practice, it takes approximately 50 seconds (20 Max Age + 15
Listening + 15 Learning) for Cat-C to take over after the failure of
Port 1/2 on Cat-B.
In some situations, switches can
detect topology changes on directly connected links and immediately
transition into the listening state without waiting max age seconds.
For example, consider Figure .
In this example, Cat-C Port-1/1
failed. Because the failure results in a loss of link on the root
port, there is no need to wait 20 seconds for the old information to
age out. Instead, Cat-C Port-1/2 immediately goes into learning mode
in an attempt to become the new root port. This has the effect of
reducing the STP convergence time from 50 seconds to 30 seconds (15
Listening + 15 Learning).
There are two key points to remember
about using the STP timers. First, do not change the default timer
values without careful consideration. When attempting to maximize
timer settings, modify only the STP timers on the root bridge
because the BPDUs contain three fields where the timer values can be
passed from the root bridge to all other bridges in the network. Consider the alternative: if every
bridge was locally configured, some bridges could work their way up
to the forwarding state before other bridges ever leave the
listening state. This chaotic approach could quickly lead to an
unstable network. By providing timer fields in the BPDUs, the single
bridge acting as the root bridge can dictate the timing parameters
for the entire bridged network.
|