Is there a driving reason as to why you are not routing in between the core and distribution layers?
While you may know this, GLBP and HSRP are FHRPs, aka First Hop Redundancy Protocols. They are generally to provide redundant or diverse first-hop gateways for a switched LAN segment. Typically in most cases the recommendation is to route between the core and distribution.
With that said, the difference between running GLBP and HSRP at different tiers of the network would depend on your topology really, and whether the failover scenarios would lead to some kind of black hole effect on traffic depending on how it occurs. I wouldn't see why you would run two FHRPs on top of eachother, though, since it implies two different L3 demarcation points.
hmmm..... That is a good question. Like a routing protocol, I would think that when you are making a redunant gateway design, you would make a decision chart that would help you decide which protocol is best. Unless there is an equivilant of redistribution for redundant gateway protocols, I would think that you would pick that protocol that is supported on all devices that would need it.
I look forward to other responses, this is an interesting question indeed.
Well I was thinking something along the lines for running 6500's at the core and like 3560's at the distribution. We can't all run 6500's at the core and distribution layer. So the 3560's do not support GLBP. They do support HSRP. So with the Cores running GLBP and dist running HSRP is there any advantage to this?
One of the things to consider is GLBP is unlike HSRP or VRRP. GLBP is an active/active forwarding scenario. You should be aware of the paths, traffic patterns and protocols. For instance, asynchronous paths don;t play well with voice and video.
Well, here is my question - why run any FHRP in the core? Core should be routed, and the distribution should route to the core.
Then, the distribution layer hosts the FHRP which services the access layer.
See what I am getting at?
When you say having "'two ISP's" , what do you mean? To properly route Internet traffic you would need to use BGP at layer 3. It all depends on your network design, but protocols like VRRP, HSRP, and GLBP really are not designed to load balance Internet traffic.
Now as far as layer 2 in the core, some newer data center technologies like virtulization (specifically things like VMotion) need layer 2. However, new technologies are emerging such as OTV in the Nexus line.
I'd suggest a peek at the design zone on cisco.com
So layer 2 traffic between access and distribution. But my Distribution layer devices are 3560's that do not support GLBP, so I will need to run them in HSRP? Or can I just run both devices and let STP take care of the loops? So I want all internet traffic to go to ISP1 and want ISP2 for lets say the IT department to manage a remote datacenter.
I may be going the wrong way on all this, but I guess I am trying to use as much equipment as I can so I dont have a device at both the core and dist doing nothing.
That is a routing concern, not a concern really for FHRP or switching, per se.
If the 3560s don't support GLBP then HSRP it is, but that still has little to do with how you route out your various ISPs. You may want to align your IT mgmt VLAN for your hosts so that the primary is closer to ISP2, but in your current topology that isn't a concern, really.
Also, HSRP runs in tandem with STP, both accomplishing different goals. STP will keep you loop free, while HSRP will allow you to create some gateway resilience via failover.
GLBP offers dynamic gateway load balancing within a given segment, where HSRP can only do manual load balancing, by splitting segments between HSRP enabled devices.
Ok so like my diagram, but instead of the Core routers going to individual ISP's they both go to one? But in that case is there an advantage to running the cores in GLBP when the distribution layer is running in HSRP?
So, there's lots of things you could do.
1) You could run a collapsed core/distribution layer.
2) You could use PBR to allow certain vlans/traffic to go over 1 ISP versus the other.
3) Design a Campus Edge switch if enough devices are available.