Mlag between x690 and x460

[HarryHeck]

I am just wondering if it is wise to do a two tier mlag between two different summit models. The x690 ISC would be 100gb. The x460 isc would only be 20gb with connection to the x690 only being 10gb.

Is it wise to have two tier mlag between systems with such a disparity in bandwidth between them?

I can connect the two seperate malgs with eaps, but the solution is overly complicated in my opinion, but it works...

[Drew_C]

A couple replies got stuck in the spam filter for some reason here... I just released them. Sorry about that!

[HarryHeck]

The 690x is the datacenter core. So server and SAN traffic will be flowing through it. The x460 will be in the datacenter as well but doing more ipmi and other traffic for devices with 1gb ports only. So the vast vast majority of traffic "should" be going over the 200gb link.

So the reason I am concerned about two tier mlag is I am not sure how pathing works. Where I am sure with the eaps design. So my fears may be unfounded.

For instance with two tier mlag lets shutdown sw01. So I loose my isc 1. I don't want a bunch of 10gb servers trying to path to the other mlag peer through sw03 and sw04. The only way I want traffic going to sw03 and sw04 if there is a host on those that a server is trying to connect from sw02 and sw01.

Looking at the two tier mlag document from extreme I think I was looking at it wrong. I thought the ISC from sw01 and sw02 would extend into sw03 and sw04. So yea I think I am going to try the two tier route. Thanks Patrick!

[Patrick_Voss]

This really depends on what is connected to these switches and will they be pushing that much bandwidth through the switches. Are the x460s going to be more edge? or distribution switches where the x690s are for the server/storage traffic?

In my opinion EAPS will provide you with more of a bottleneck because you will be losing one uplink. You could remove the ISC on the x460s and have 40gb split up to the x690s. This will remove some of the redundancy on the x460s but increase the bandwidth to the x690s.

[HarryHeck]

Here you go. They will not be providing any routing or layer 3 service. I am just concerned about putting these 4 switches together in a 2 tier mlag. The 690 is 48 ports of 10gb and the x460 is 24 ports 1 gb ports. With the x690 having 200 gb between them and the x460 on 20gb there is a huge disparity between the separate isc bandwidth capabilities. If the pathing between the x690 would always take the 200gb link I would feel better about this setup.

My other design with eaps is similar below except the two isc and not tiered together. I have to use eaps to prevent the loops which adds complexity.