Chapter 7. Nortel Networks L2/3 GbESM configuration and network integration 109

Reasons for choosing a Layer 3 design

The reasons for choosing a Layer 3 design for blade server switching are outlined below. It is
worth noting that in most discussions of network architecture, the blade servers would be
connected to a so-called server access switch. This is a switch at the edge of the network
which connects directly to servers as opposed to client (desktop, mobile computer)
computers. Most discussions of network architecture recommend that server access switches
be Layer 3 devices.
򐂰Layer 3 switching keeps more traffic within the BladeCenter chassis.
In any design where the blade servers are assigned to multiple VLANs, routing is required
for servers which are not on the same VLAN to communicate with each other. (An
example of such a design would be using blade servers as WebSphere® Web and
application servers and placing the Web servers on one VLAN and the application servers
on a different VLAN.)
With a Layer 3 configuration on the GbESM, servers on different VLANs can communicate
via the switch module which is inside the BladeCenter chassis. If only Layer 2 switching is
used within the chassis, then traffic between, for example, WebSphere Web and
application servers would leave the chassis on the external links and flow through one or
more external devices until it reached a Layer 3 switch (router). It would then flow through
one or more additional devices until it returned to the IBMEserver BladeCenter,
crossing the external links for a second time.
The benefits of keeping traffic within the chassis include greater security - because there
are no patch panels which can be tampered with - and lower latency, since the traffic
traverses a smaller number of switches between source and destination.
򐂰Layer 3 switching allows more efficient use of external connections
The key issue here is that use of Layer 3 allows the network to run without the use of
Spanning Tree Protocol (STP). Spanning tree works by blocking links which would create
a topology which includes a loop; connections from a GbESM to two or more upstream
switches which are connected to each other fall into this category. The consequence is
that up to half of the links from the GbESM would be blocked during normal operations
and would not carry traffic. Only if the active link(s) failed would the blocked links be used
to carry traffic.
Layer 3 routing not only allows all of the uplinks to be active but also allows the GbESM to
send traffic to a given destination on the best path to that destination.
7.8.1 Layer 3 sample configuration with static routing and VRRP
This example is an extension to the basic configuration described in 7.6, “Basic Layer 2 entry
topology” on page69, providing a mesh topology between the two GbESM switches and the
upstream Core switches. Trunking (link aggregation) is used in this configuration in the same
way it was used in the advanced Layer 2 configurations shown in 7.7, “Advanced Layer 2
topology sample configurations” on page80.
See Figure 7-6 on page111 for a diagram of the topology used in this example.
VLAN 35 is configured and connects both of the GbESM switches to Core switch 1 (address
.245). VLAN 46 is configured to connect both of the GbESMs to Core switch 2 (address .246).
This configuration uses VRRP and hot standby to provide high availability.