80 Nortel Networks L2/3 Ethernet Switch Module for IBM Eserver BladeCenter

7.6.2 Basic topology conclusions

Testing for this configuration involved pulling cables from the g0/1 and g0/2 ports of Core1
while pings were running. Removing the cables one by one shows that first the trunk
redundancy switches the link. After removing the second cable, the trunk failover brings down
the INT ports of GbESM_1. If the Nortel switch is being monitored, Figure 7-4 illustrates what
you see. After 1 minute and 30 seconds of pulling the second cable of the trunk, we
reattached both cables to Core1.
Figure 7-4 INT ports blocked during a trunk failover test
We also noticed in testing this configuration that there was no difference in the failover
behavior between LACP and a static configured trunk. LACP is not shown as configured in
the examples because the current generally available code release does not support trunk
failover with LACP.
One ping was lost on the first cable pull, while only three were lost on the second when the
failover occurred. At most, only three pings were lost when the cables were reattached and
failback occurred.
7.7 Advanced Layer 2 topology sample configurations
This example is an extension to the basic configuration described in 7.6, “Basic Layer 2 entry
topology” on page69, because cross connections between the GbESMs and the upstream
switches are established now. See Figure7-5 on page 81 for the topology used in this
example. The crosslinks increase the redundancy and provide more flexibility regarding the
configuration of the BladeCenter components (for example, trunk failover is not needed for
full redundancy) but it costs the need for loop prevention. This is commonly reached through
the usage of a Spanning Tree Protocol, blocking dedicated ports to break the loop on Layer 2.
Another approach to break the loop, based on the Layer 3 capabilities of the GbESMs, is
shown in 7.8, “Layer 3 topology sample configurations” on page108. The initial configuration
for all the switches in the following examples are the same as those in 7.6, “Basic Layer 2
entry topology” on page 69, except that IEEE 802.ad dynamic port aggregation (LACP) is
used on all the upstream connections rather than the static Etherchannel.
Note: LACP with trunk failover is a feature to be included in an upcoming release. During
this experiment a early version of this code was tested and the feature does work.