Best way to connect access with distribution with limited options

Hello, Community! Thinking what would be the most effective way to connect the devices shown in the below topology (two X460s in the access layer connecting to one X670 in the distribution).
1. One approach would be to interconnect them via EAPS domain. However, if this would scale of up to 20 domains on the X670 (up to 40 access layer switches), wouldn't that make some operating / managing issues ?
2. Other option - stacking the X460s in the daisy chain (only option because of the 10Gb uplinks) is not recommended, because the loss of the link between them makes for all sorts of troubles (double masters, etc.).
3. Maybe you would recommend some other solution that would be more resilient and scallable ?

6 replies

Userlevel 6

1. If it's old X460 - you can use SummitStack module for sterling stack and XGM3-2SF module for uplink
2. If it's X460-G2 - yu can use VIM SummitStack module for stacking and 10G ports on the front of switch for uplink
3. If there is no SummitStack module you can use 10G ports for stacking and 8 1G ports for LAG uplink - not really a good solution, but it has the right to life.
4. You can use M-LAG, X460 with ISC (few 1G ports) and X670 with LAG
Thanks, AlexandrP!
The X460s are with XGM3-2SF module each (other slot is empty). Regarding your suggested option 4, if we configure MLAG, with a couple 1Gb ISC ports between X460s, wouldn't that be sort of a bottleneck for traffic, in case one 10Gb link fails ?
What is your opinion regarding 20 EAPS domains on one distribution switch ? Would that be a feasible option ?
Userlevel 6
About M-LAG good redundancy you can see here -
About 20 EAPS domains - Summit series switches have limitation up to 32 EAPS domains (you can look this limit in RelNotes for EXOS).
Userlevel 2
I prefer to have option 1, maybe with some optimization of X460 number in each domain if possible.
Userlevel 6

Any and all alternatives mentioned suffer the same BIG problem. A single X670 connecting 40 X460 to the rest of the network is a HUGE single point of failure.

I would add a second X670, configure MLAG and dual home each X460 through their 10G ports.
I think we will be going with the EAPS. Thanks for the opinions, guys!


I'm aware and couldn't agree more with you regarding the single point of failure in the core. The intended topology was fully redundant. However, due to some reasons the client chose to split the architecture. We made sure that they're fully aware of the situation and potential consequences and i'm trying to make the most with what we have.