IBM L2/3 Switch User Manual


 
90 Nortel Networks L2/3 Ethernet Switch Module for IBM Eserver BladeCenter
Design or topology remarks
Although this topology is very robust and offers high redundancy, newer topologies do exist.
The drawback of this topology is the convergence time after a connection-loss or a
switch-failure. Our tests showed that as long as only one link of an aggregated port group is
going down, the switchover time is around 1 second, which is fairly good. In contrast, when
the whole aggregation (both links) or one of the switches itself go down, it takes 30 seconds
to converge, which might be unacceptable for some applications. This convergence time is
the natural limitation of IEEE 802.1D Spanning Tree: 15 seconds Listening-state and 15
seconds Learning-state before a port reaches Forwarding state.
It is good that in this topology we do not have to rely on trunk failover, which will not be
supported on LACP trunks until the next (1.1) software release. This topology also does not
need BladeServer NIC Teaming, although the BASP Teaming function should be considered
with regards to its high availability capabilities.
As seen in 7.6, “Basic Layer 2 entry topology” on page 69, a better convergence time is
reached with our basic topology but only by interacting with trunk failover and active or
standby NIC teaming. Other ways to reduce the time that is required to recover from a
topology change include recent enhancements to the Spanning Tree Protocol, such as
Rapid Spanning Tree (IEEE 801.1w) , which is discussed in the next section or, alternatively,
a Layer 3-based solution, which is discussed in 7.8, “Layer 3 topology sample configurations”
on page 108.
Full configuration snapshots
Example 7-13 shows the output of /c/dump for the GbESM_1, and Example 7-14 on page 92
shows the output of show running-conf from the Core1 Cisco3560 Switch. See the
differences to the configurations of GbESM_2 and Core2 as comments in the relevant lines
respectively.
The configurations are valid for copper and fiber infrastructure, it makes no difference to the
GbESM. For the uplink switches the difference is, that for fiber connections the GBIC ports
G0/25 and G0/26 have been used as IEEE 802.1Q tagged ports (not aggregated) and G0/27
as interconnection of the Cisco Switches.
Example 7-13 Advanced Layer 2 Topology with PVST configuration for the GbESMs
/* Version 1.0.1.6, Base MAC address 00:11:f9:36:b7:00
/* GbESM_1
/c/sys
hprompt ena
/c/sys/access/user/uid 1
name "USERID"
pswd "cbd31b7a4b020a2a86b6f2b388b2a9d8e8b4271b97d91dc22045f70228a24127"
ena
cos admin
/c/sys/ssnmp
name "GbESM_1”
/* name "GbESM_2" on GbESM_2
/c/port INT1
pvid 20
Important: When looking at the configuration commands presented, commands toward
the left margin are for Core1 or GbESM_1 switches. Where the GbESM_2 or Core2 switch
configurations differ, the commands presented within comment syntax (/* for GbESM, !
for Cisco, # for Extreme) are for the counterpart switch of the same type, GbESM_2 or
Core2.