Three-way SMLT/vIST

[Nicolas_Melay]

I have SMLT/vIST working between two VSP7400.
I need to create antoher SMLT between one of them and a third VSP7400.
There's no way to make this happen, right ?

[Jave · Accepted solution]

Hi Nico,

No, for me, it's impossible, one VOSS can by design only be member of one single vIST.

Regards,

Jave

[TQU · Accepted solution]

Hi Nicolas,

No way, cluster SMLT/vIST are only by pair.

Why can't you attach the device on first cluster ?

regards,

TQU

[Nicolas_Melay]

Thanks Paul (and others) for the confirmation.
Yes, ESXi actually uses standalone links more efficiently than LAGs.
But we also have NetApp filers that need bandwidth more than anything else, and I really need LACP here.
Like I said, I'll switch EXOS stacks to STP, and probably submit a RFE for this specific case.

[Nicolas_Melay]

Sorry, I guess I did not clearly express what I want to to.
My goal here is to connect other devices (EXOS stacks and servers) to VSP7400's.

The current network is built around 2 VSPs, each in a different server room.
Remote EXOS stacks are attached to them using SMLT, with a fiber connection to each room.
A few servers, located in either room, are also attached to both the local and the remote VSP (over room-to-room fiber interconnect), again using SMLT. Of course this does not scale, as we end up "wasting" a lot of fiber links in this interconnect.

We planned to install a second VSP7400 in each room, in order to get local redundancy for servers, and get rid of the remote links.
(See the diagram below, maybe if it can explain this better than words.)
But now I realize I can't both have SMLT trunks from VSPs in different rooms, and other SMLT trunks from VSPs in the same room.
Seems like by best option will be to fallback to STP for EXOS stacks, and lose half the bandwidth.

[EXTR_Paul · Accepted solution]

i see your image now

the SMLT and LACP links must terminate on the same vIST cluster. So in your example you can't have switch stacks going to a vIST cluster spanned across two DC's and then support MLAG to the servers in the same DC.

in your example core A1 and B1 must be in the vist peers for each other. But that would break your LAGs to your servers.

the only way you could do this is research what kind of NIC teaming your servers support. Some servers like ESXi and Microsoft have an L2 teaming feature that doesn't require a LACP LAG on the switch side. you only need to configure access or tagged ports.

[EXTR_Paul]

The vIST MC-LAG solution allows for two switches in a cluster.

If you wished to connect a 3rd VSP to the dual viST cluster you would just use SPBm with NNI's. Don't use viST/SMLT.