03-06-2014 01:33 PM - edited 03-07-2019 06:34 PM
Our environment is set up in a ring topology.
The outside 6500's is the ring.
The two 6500's in the middle are at our second campus and are introduced into our main campus enviroment as shows.
We have an issue that traffic from our second campus prefers one connection over another and all metrics are basically the same. We use vlan's for everything and according to eigrp the bandwidth is shown as 1G, depsite some phsyical interfaces are 10G.
In the enviroment, the 6500 to the far right, mchh, shows traffic utilization around 70-85%.
We are working an issue with a video endpoint that has poor service. If we move the device to another area on our environment and see the traffic go accross a different 6500, not the mchh 6500, our service is not impacted. So we can see where the issue is at.
If we move a connection from the dow01-cr-02 6500 to dow01-cr-03, our problem could go away for this issue but it does not solve for the route selection and why the hh 6500 is preferred for just about everything.
I am considering adding variance to the environment, but unsure where to introduce it and what it could do to other traffic.
Any thoughts?
Any ideas of what else to look at?
03-06-2014 04:32 PM
Rick
Just as a follow up.
I am not saying that is the only solution. You could look to use variance but the majority of traffic would still be sent via the mchh switch because that is the best path. But it may be a possible solution to balance the traffic more evenly.
You can also use delay on certain interfaces which would affect all routes received over that path or offset lists to prefer certain subnets one way and certain subnets the other.
Finally it does depend on where your 10Gbps links are in relation to all the switches.
All of the above manipulation obviously introduces complexity though so it is a tradeoff really.
Jon
03-10-2014 04:35 AM
Rick
Do you mean most traffic between all vlans goes via the mchh 6500 ?
If so then assuming all the interconnects are seen by EIGRP as the same bandwidth and delay it would do ie. if you count the L3 hops then you can see that from dow01-cr02 to get to the majority of the vlans on the other switches the shortest path is via the mchh 6500.
The only exception to this are the vlans on mclc21-cr-01 which from dow01-cr02's perspective has equal cost paths via mchh and dow01-cr03.
So i would expect the mchh switch to be routing most of the traffic to all the other vlans.
If on dow01-cr02 you do -
1) pick a vlan on mcuh21-cr01 and do a "sh ip route" for the vlan IP subnet. I would expect to see one route pointing to the mchh 6500
2) pick a vlan on mclc21-cr01 and do a "sh ip route" for the vlan IP subnet. I would expect equal cost paths via mchh and via dow01-cr03
Again it does depend on EIGRP seeing all connections as the same bandwidth and delay.
If the above is the case then the simplest solution, if possible is to move all the vlans from dow01-cr02 to dow01-cr03 because that sees more equal cost paths and so traffic would be more evenly balanced.
But this is assuming all metrics are indeed the same and that you could relocate those vlans.
Edit - it also assumes that traffic between all vlans is relatively evenly spread.
Jon
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide