Redundent network design question

[Adam Watts]

Hi,

I have an idea on how to add some redundancy in to my existing network and would like some feedback on if it will work.

We currently have network which is currently going through a design change and as part of this my boss is replacing most of our stock of catalyst 3524 and 3550's with catalyst 3750 stacks. Also we are adding some 4500 series switches at the distribution layer these then link back to our main core 6509 switches via 10Ge links. In the core it has been decided to have no Layer2 traffic at all so all links are running as OSPF links and all gateway address are setup on the distribution switches with traffic then routed through the network. What I want to do is to put some redundancy into the distribution layer as at the moment if the 4500 was to fail we would loose the stacks and the attached end users. but due to financial and space restrictions we can't duplicate the 4500s. the idea I had was to put another distribution switch in another location and connect the stacks back to the secondary switch with our existing fibre network, then setting this second switch with low priority with HSRP and Spanning tree, so that if one of the primary distribution switches went out the secondary would kick in. The secondary distribution switch would need to be a secondary switch for several distribution groups.

I've attached a diagram to try and illustrate what I am after. with the possible setup for the third (secondary) distribution switch. the diagram only shows two distribution and access groups but I have 4 and will be adding a couple more in the future.

if anyone has any ideas on this or wants more information then please get in touch.

Adam

[lgijssel · Accepted solution]

The 4500 series has a very high stability anyway and you will gain little by trying to improve it. Perhaps you should rather leave the design as it is, i.e. without redundancy. My opinion is that this kind of hap-half measures (excuse me for the expression) are likely to cause more troubles than they will solve. They will make your network harder to maintain and troubleshoot.

Regards,

Leo

[lamav]

Why not create layer 2 trunks between each distribution switch and each access switch stack? This way, if one of the distribution switches goes down, the other will have full connectivity to all the access switches. No?

[albert.remo]

Hi Adam!

As I see in your setup, I think Dist3 is no longer necessary for it might make the STP recalculation during failure more complicated. Dist1 and Dist2 will be well enough to provide redundancy for each other. Then connect the 3750 stacks to both Dist1 and Dist2. This Core and Distribution Layer design would already give you a very high percentage of availability. Scalability will now depend on the Supervisor you will be using. Try to go for a Supervisor V-10GE which has a Switching Capacity and throughput of 136Gbps. Just to add, make sure not to connect both primary and secondary uplink to the same switch on the stack.

Hope this helps!

Regards,

Albert