Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 
Community Member

STP design advice



8 x 3750's - Distribution

2x 6500's (Layer 2 & 3) - Core

I was planning on implementing two Catalyst 6500 switch\routers which would feed 8 3750's via fibre. Each 3750 has a feed from both 6500's. The 3750's are directly connected to our Server cluster.

Both redundancy and bandwidth are a high priority, however I can't seem to find a design that will allow each switch to have a feed from each of the core 6500's and still use the maximum 4gb of bandwidth because of STP. The servers are using Microsoft cluster services so layer 2 connectivity needs to be available between each server (ruling out layer 3 routing).

Is it possible to avoid switching loops and still get maximum throughput? Any help would be appreciated :)

VIP Purple

Re: STP design advice


I think your best option would be to implement VLAN Load Balancing. The idea is basically to distribute your VLANs over both trunks by setting the port priority for each VLAN. Let's say you want your VLAN 1 to use GigabitEthernet0/1, and your VLAN 2 to use GigabitEthernet0/2. The configuration of the interfaces would look like this:

interface GigabitEthernet0/1

spanning-tree vlan 1 port-priority 64


interface GigabitEthernet0/2

spanning-tree vlan 2 port-priority 64

The default value is 128, so as long as you specify a lower value on an interface, the VLAN will choose that link.

Have a look at the document below as well:

VLAN Load Balancing Between Trunks Using the Spanning-Tree Protocol Port Priority



Community Member

Re: STP design advice

You can redesign your network in such a way that you don't have to rely on STP at all. This can be achieved by DIS-allowing VLAN/Layer-2 loop. There are 2 options to do this. Pls refer to the attached diagram (Visio 2003 format).

In the top diagram, you can see that the VLANs are not looped. With this topology, STP convergence won't come into play in case of an uplink failure.

The bottom diagram shows a somewhat "radical" idea (Cisco is starting to sell this idea quite rigorously, actually). This second topology basically requires configuring your 3750s as Layer-3 switches (standard image comes with basic Layer-3), and make them the default gateway for the servers. You can then run either RIP or even multiple static routes between the 3750s and the 6500s, which should be able to utilise both uplinks, so long as they have equal costs. If the costs are somewhat different, you can use EIGRP that can do unequal cost load balancing. However, this will require you to upgrade your 3750 IOS to enhanced image.


CreatePlease to create content