Solved: EIGRP timers and bandwidth over Metro-Ethernet?

Nathan Eger · ‎08-29-2008

Hello All!

Question about configuring EIGRP routing over a Metro-Ethernet connection:

Setup:

SIDE A

4507 100mb int. connected to Metro-E rate-limited port of 20 mb

SIDE B

3570 100 mb int connected to Metro-E rate-limited port of 20 mb.

EIGRP AS of 10 configured on both routers

3750 configured as a stub with summary summary addresses advertised to neighbor.

Question:

Any recommended timers for hello and holdtimes?

Any recommendations about bandwidth settings for usage of EIGRP of the link?

THANKS IN ADANCE

Nate

Giuseppe Larosa · ‎08-29-2008

Hello Nathan,

I would use default timers.

As a refinement you can set the bandwidth command to the actual rate of 20000 kbps.

In order to be sure to avoid losting adjacencies during heavy load you may want to use modular QoS to provide some resouces to EIGRP and to shape all other traffic to something less of 20 Mbps.

This is to avoid cases where you alternate huge traffic peaks with nothing and having EIGRP up and downs if their packets are discarded in the MAN.

If you use modular QoS setting interface bandwidth to 20000 kbps is required to use commands that use percentages.

Hope to help

Giuseppe

View solution in original post

Giuseppe Larosa · ‎08-29-2008

Hello Natan,

the eigrp bandwidth command is useful when dealing with slow links or in point-to-multipoint WAN where you set a bandwidth lower then real one and then allow eigrp to use more then the declared BW.

with a 20 Mbps pipe and having EIGRP a 1 packet update window (don't send next update until I receive ACK for the previous one)I don't expect any issue on this.

EIGRP sends updates only when something changes ( no full updates like RIP, no data structure periodic refresh every 30/15 minutes like OSPF and IS-IS).

So I wouldn't use the ip eigrp bandwidth command.

Thanks for your remarks of first post.

Hope to help

Giuseppe

View solution in original post

Giuseppe Larosa · ‎08-29-2008

Hello Nathan,

I would use default timers.

As a refinement you can set the bandwidth command to the actual rate of 20000 kbps.

In order to be sure to avoid losting adjacencies during heavy load you may want to use modular QoS to provide some resouces to EIGRP and to shape all other traffic to something less of 20 Mbps.

This is to avoid cases where you alternate huge traffic peaks with nothing and having EIGRP up and downs if their packets are discarded in the MAN.

If you use modular QoS setting interface bandwidth to 20000 kbps is required to use commands that use percentages.

Hope to help

Giuseppe

Nathan Eger · ‎08-29-2008

Thanks for your reply. I will leave the timers alone, set the interface bandwidth settings, and configure QoS.

What about the use of the eigrp bandwidth command? Does it make sense to limit eigrp usage of the link to less than 50%? Obviosly, some perfromance metrics on the link usage would be needed to accurately anwer this question, but do you have a rule of thumb?

Thanks again

Nate

Giuseppe Larosa · ‎08-29-2008

Hello Natan,

the eigrp bandwidth command is useful when dealing with slow links or in point-to-multipoint WAN where you set a bandwidth lower then real one and then allow eigrp to use more then the declared BW.

with a 20 Mbps pipe and having EIGRP a 1 packet update window (don't send next update until I receive ACK for the previous one)I don't expect any issue on this.

EIGRP sends updates only when something changes ( no full updates like RIP, no data structure periodic refresh every 30/15 minutes like OSPF and IS-IS).

So I wouldn't use the ip eigrp bandwidth command.

Thanks for your remarks of first post.

Hope to help

Giuseppe