Redundant layer 3 switches

[Bart Kersten]

Hi,

I'm currently busy with designing a redundant network for in a datacenter. I know you could make 2 layer 3 switches redundant with Cisco Stackwise. I also know it can be done with HSRP.

But, can it be done any other way? with this i mean is there another way two make 2 layer 3 switches, lets say 2 3560's and make them funtion as 1 switch? I experiment a little bit so far and couldnt come up with a solution other then Stacking or HSRP.

Thanks in advance,

Bart

[Peter Paluch]

Hello Bart,

There is a huge difference between the StackWise and HSRP. As you probably know, the StackWise technology is usable only on 3750 series Catalysts, and it allows them to appear and operate as a single switch - they have a single configuration, single MAC address table contents, a single set of own MAC and IP addresses - for all practical purposes, the stack of switches interconnected with StackWise behaves like a single unit. The HSRP merely makes the two switches appear as a single IP router to hosts in attached VLANs but the switches still operate separately and independently. Therefore, you cannot really compare StackWise to HSRP because the redundancy is performed at vastly different levels and therefore results in major differences in the overall behavior.

The StackWise is not available for other switch series than 3750, so specifically for 3560, the StackWise is not an option. The 2960-S Catalysts have a similar feature called the FlexStack, but that's about it.

Chassis-based Catalysts like 4500 or 6500 are inherently redundant and may be equipped with redundant supervisors, so in their case, the redundancy is provided by the presence of multiple supervisors in a single chassis. In addition, the 6500 and perhaps soon the 4500-E support the Virtual Switching System (VSS) allowing two 6500 units behave like a single 6500 unit.

It basically depends on what kind of redundancy you are looking for. Basic connectivity redundancy is achieved by STP, EtherChannels and routing protocols, and perhaps first hop redundancy protocols like HSRP, VRRP, or GLBP. More complicated scenarios like cross-stack EtherChannels where different links in a single EtherChannel bundle are connected to different switches in a single stack, require the presence of a StackWise interconnect or a chassis-based switch. You see... we need to know more about the requirements on redundancy you have.

Best regards,

Peter

[Bart Kersten]

Hi Peter thanks for your reply,

I will attach a design of what i had in thought.

It think its a fairly typical design. like i said i alerady have tried to expermiment with the design in packet tracer. I used OSPF to route al the networks and used trunks and etherchannels for the vlans. I know a 3560 doesnt support stacking, but i didnt know as much of HSRP thanks for clearing that up.

In this scenario i want the 2 routers too function as 1 gateway for all traffic but i still want them to be redundant, this is achieved by HSRP right?

the redundant links between the routers and switches are just normal layer 3 links with ip adresses assigned to them. the links between the layer 2 and layer 3 switches are trunks and etherchannels. (correct me if this isnt the right way to go )

As routing protocol i am using OSPF, i have OSPF configured on both routers and switches.

Till this far all went well.

How do i create redundancy the correct way from to layer 2 to layer 3 switches, I probably have to make etherchannels with trunks between them? And what are the correct configurations of the layer 3 switches? do i have to give the vlans on both switches an IP so that if on goes down all traffic is routed to the other IP of the same vlan on the layer 3 switch? what i mean by this is to create 2 gateways per vlan.

Any suggestions are welcome!