GLBP (Gateway Load Balancing Protocol)

This topic is to discuss the following lesson:


Hi Rene,
How many AVFs supported by GLBP?
From the GLBP Virtual mac address 0007.b400.XXYY , I can found 256 Group , and 256 AVF.
Is it supports 256 AVF?


Hi Srini,

I think it depends on the hardware platform. I just took a look around and the supervisor 720 supports up to 1024 groups.


1 Like

where do i choose an ip address from ? I want to implement GLBP between 2 core routers -

BGP router ( ISP )
Core router ____GLBP ______Core router
Layer 3 switch
Fire wall
Layer 3 switch ( access level , supporting VLANS ) subnet

Hi Aaron,

GLBP is configured for each subnet so if you want redundancy within the VLAN that has subnet then that’s where you’ll need to pick an IP address from.


Hi Rene

I tried that but then the core router was advertising that as connected network , yet the other nodes were showing that as a advertised network learnt through EIGRP and then I was unable to send pings to the virtual IP or the GLBP interfaces which were configured with an IP address from that VLAN subnet ??

its only redundancy for the core routers incase one fails and also to provide some sort of load balancing as there are two different sites that replicate each other as per the diagram ??

Hi Aaron,

If you run EIGRP between your Core routers and L3 switch then why do you want to use GLBP here? If one of the two core routers fail, EIGRP will use another path.

GLBP (or HSRP or VRRP) is normally used for end devices that have to use a default gateway…there’s no way for these devices to pick another default gateway in case the first one fails :slight_smile:


1 Like


what can i use for load balancing at the core level ? so you say there is no need for any type of redundancy protocols what so ever ??

thank you for your time !! i want to get this right


Hi Aaron,

Routing protocols like EIGRP or OSPF provide load balancing if the metric is equal. If your core routers are advertising routes with the same metric to your L3 switch then it will install two entries in its routing table.



could you show me a copy of the config that I would use for load balancing and redundancy at the core level if I am using /27 subnet for the link between the two and also for the eigrp interfaces that are connected to each other ??

thank you

Hi Aaron,

It’s basically the same example as here:

Just imagine that router Preston is your ISP router and Carl/Ann are the core routers. KingKong is the L3 switch.


Hi. Building upon the present GLBP configuration example how would you set it up such that switch A is the AFV for roughly 75% of the outgoing packets and switchB is the AFV for roughly 25% of the outgoing packets?

Many thanks.

Hi Thomas,

You could try 75 for SWA and 25 for SWB, that means that SWA will be used 3x more often than SWB. 150/50, 100/33, 200/66 should achieve the same.

Keep in mind that this is the number of “hosts” served, if all your hosts use the same applications then it should be equally balanced. If one host is running 200mbps of torrent traffic and another host is using telnet then one of the switches will still serve 99% of the traffic :slight_smile:


1 Like

Dear Rene,

Just curious question. Why HSRP v2 and GLBP are using same multicast IP address though they used different port number?

Why not use different multicast IP address?




1 Like

Dear Zaw,

No idea…:slight_smile: If I had to take a wild guess I’d say they did it since the 224.0.0.x space is limited. It doesn’t matter though, they use different port numbers.


1 Like

If there is only one host that needs to get out, will glbp still load balances the traffic? What about if there is active/standby pair of Firewall that needs to get out?
As in the example shown above, two different host are showing two different destination virtual mac, I am wondering in cases I mentioned above. Will only one AVF forwards traffic as AVG is seeing traffic coming from only one source mac address ?

This depends on the GLBP load balancing method and how “chatty” the network is.

If you use a non-default GLBP load-balancing method of host-dependent, then the traffic will never be load-balanced in your case.

For the other two methods, it is possible the traffic will be “load-balanced” but only over a very long period of time. Once a host learns about a gateway’s MAC address via ARP, there will be no chance of load-balancing until that host has to re-ARP for the gateway’s MAC again. In order for a computer to re-ARP, there has to be no traffic sent to its gateway for more than the ARP timeout period.

So, assuming you are using either Round-Robin (the default) or Weighted for GLBP, AND you don’t have much traffic at all from the host in question leaving the subnet, then yes, could you load balance in theory. But in reality, no. :slight_smile:

Thanks Andrew for quick response!! That answers my questions perfectly! Awesome !

Great :slight_smile: