first of all, they way u thinking about to configure ur network with GLB is very good and better than HSRP

but, u have a problem with the way that the switches connected

u need two uplink connection from the access switches to the 4500 switches because if u have one uplink and one of the 4500 gos downs

u will not be able to get the beneft of the GLBP becuase all the switches connected to the

one gone down will be disscontcted fron the other 4500 and from the server farm

so just make ur link between the switches i mean the uplinks redundant

and they should be connected the same way as the servers connected to them

so they can use the GLBP virtual IP address and u can achive redandancy and loadbalncing

good luck

please, if helpful Rate

ok question with your reply. I understand that if i lose a 4500 half of the switches would go down but the other half would stay up and be able to access the servers right? Unfortunately the 3560's are spread out across the entire county and i can only have one uplink to my core although maybe in the future i can get a few sites with dual uplinks. This is the best possible redundancy and LB method i can come up with. half of the switches on one 4506 and the other half on the other 4506. So besides losing half of the switches during an outage are there any other concerns you see? One more thing to add, the links between the 3560s and the 4506s will be layer 3. so no STP.

if u are unable to make dual links

i think u dont need to make exra config

like HSRP or BLBP

becuase, if u gonna lose that half in case of one of the 4500 gos down u will not get the benifit from those protocols

just keep it two hlfs ur network and as simple as possiable

unless u can find a way to get a redundant link

once u get this link go ahead and make the GLBP

this is what i see thorugh what u have in ur case

and its up to u

good luck

ok thanks. i think i will go ahead and run glbp. i want it in place in case we get redundant uplinks. the building the data center is in now can have about 10 switches with dual uplinks, but i have about 20 more spread out that currently don't.

good luck :)

if helpful rate

Hi,

Just some notes.

The servers have dual connectionsn to the network. Genrally one will be active (sometimes both) but if you use GLBP gateway then there is every potential for them to hit the Left L3 switch go across the trunk to the Right L3 switch as it default GLBP gateway, then back to the Left L3 switch down tree to reach the destaination.

So where you potentially share the load between switches you acutally increase the load on the inter switch links.

With HSRP you can manually tune the network for match L2 forwarding so under normal operation you have know which path traffic should go. Using GLBP you lose this deterministic approach.

The question before deploying GLBP is what are you load balancing?

Is is out going traffic becasue of bandwidth issues?

Is it load on the L3 router?

My view of GLBP is that it is best used in small office (branch)network with multiple WAN connections where client would be directly connect to VLAN with the WAN routers and we would like to balance outgoing traffic to the WAN.

The main reason is that the L2 forwarding capacity (1gb min normally) is well belows the WAN threashold 2 x 100mb links or less so L2 forwarding does not really come into the equation.

When trying to load balance xGB links then L2 forwarding path is a big concern.

hope this helps.

Interesting. What my over all goal is redundancy in the data center. We have a campus network spread out over a county, all gig Ethernet links. Currently it is all layer 2 trunks and I am spanning multiple vlans over multiple switches. I want to change that. Like i stated in the original post, I will have half the switches connected to one 4506 and the other half connected to the second 4506 all via layer 3 links. I chose to go with GLBP in the data center on 2 4506s rather than HSRP for the load balancing feature. It's hard to go with HSRP when GLBP is pretty much the same thing and it load balances. Maybe I have a distorted perspective on GLBP. Either way i just need redundancy. I have a new 4506 coming in soon. My core will be 2 4506's and no distribution switches, but a routed access layer with 3560's. Eventually i will get a 6500 for a true core but my budget won't allow this year. The long term plan is to make these 2 4506s distribution switches and a 6500 as the core

as i mentioned above

wihtout redundant uplink not binefit from GLBP or HSRP

by the way

u gonna go to routed access switches

if u had redundant links

u could have done the best tpology

which is based on routing protocol loadbalncing useing equalpath and also in case of any link down the redunndant link will take over with 0 convergenace time !!

which is more reliable

thank you

So you are saying that the servers will not benefit from HSRP either? If i lose a 4506 and half of the access switches go down the other half would stay up right? and be able to access the servers? I know i can't get the full benefit of glbp or hsrp right now, but i hope in the future to get full redundant uplinks. AT least it would provide some sort of failover.

If we had the budget every device on my network would have 2 gig fiber uplinks. We have a hand full of switches with 2 uplinks, but the majority do not. We have dark fiber from an ISP. Routed access switches is what I'm moving to now and my 4506s will be used as distribution/core for now until i can get a 6500. So basically everything connects back to the 4506's which are in the data center with the servers.

once u get two uplinks for each switch

then u can get the binefit from GLBP, HSRP or ROUTED ACCESS LAYER

with one uplink no benifit at all

any Dist or core layer switch fail all switches connected to it will be isolated

hop this helpful