01-05-2013 07:00 AM
Can some body please let me know if a VCS fabric can span 60 VDX 6720 and VDX 6730 switches, I've got 2 MLXe-16's in the core and each VDX is dual homed to
the MLXe's.. or if its not possible can multiple islands of Fabrics exists and would they be able to communication with each other ? would much appreciate an input ASAP
01-05-2013 08:35 AM
NOS 3.0.1 Release Notes states that the maximum members a fabric can have is 24.
Although, this is the official number tested and validated in R&D labs, we were told this figure can/will increase with time.
As per your other request, AFAIK it's not actually possible for multiple islands of fabrics to communicate with each other without an intermediate device.
01-05-2013 08:55 AM
Thanks ybzahr, since I have the 2 MLXe-16's on the core cannot I use the same for the routing towards different fabrics ?..
In my context I have 30 Racks out of this 20 Racks would be deployed with (each rack) with 2x VDX 6710's in while on the remaining 10 Rack, each rack would have 2x of VDX 6730's and will have storage devices with FC and FCoE support.
The issue I am facing is how to bridge the connectivity between the 20 racks with VDX 6710's and the 10 Racks with VDX 6730's.. any insight would be appreciated please..
01-05-2013 09:09 AM
Well, you pointed out the problem : it is about bridging.
If you really want to build an ethernet fabric, then you need to connect the VDX switchs toguether in some sort of topology (star, ring, etc.) that is totally up to you.
Problem you will hit are :
- scalability : 24 members maximum / per fabric,
- economical : fabric licences + FCoE licences,
I guess in case you are describing, you could bridge all your vlans from your TOR switchs through your MLX core boxes. You would then have 30 entirely independent fabrics with your MLX core boxes running as big switchs.
==> Keep in mind your TOR will never act as one big fabric (60 members) as they are not directly connected.
Other option, you could always use L3 routing from you core boxes down to your TOR switchs using the newly introduced routing features in NOS 3.0.
01-05-2013 09:39 AM
Thanks ybzhar, my concerns is how would the Servers on the VDX 6720's will access the Storage on VDX 6730's which in this case would be in a difference fabric... guess after all the fabric thing is not sweet as it sounds there are scalability issues no matter how hyper it is out there....
Could you please let me know if two independent fabrics can talk to each other and exchange FCoE Traffic ?
01-05-2013 10:40 AM
--->>>.....if two independent fabrics can talk to each other......
You question is a question itself.
can you please explain me what you mean with:
"two independent" ..../.... "to each other"
is for me a bit confused.
01-05-2013 05:22 PM
in the case of a fabric with two or more members, FCoE initiators can reach distant FCoE targets if all members have the proper licences (ethernet fabric + FCoE storage).
in the case of separate fabrics, i don't think it is possible without some kind of intermediat box.
At this point, if you have distant storage to reach, i'd suggest you form one or two big fabrics but it'll cost extra $$$ and you will hit the limit of 24 members / fabrics.
all systems have their limitations, that is true for all vendors... That's something your local SE can help you prevent by designing a solution with you ;-)
01-05-2013 07:01 PM
In my context there would be :
1. 20 Nos of Racks (A) each with 2 Nos of VDX 6720's dual homed via 10GigE to the 2 MLXe-16's (core)
2. 10 Nos of Racks (B) each with 2 Nos of VDX 6730's dual homed via 10GigE to the 2 MLXe-16's (core)
It is approximated that there would be a max of 24 servers per rack.. the issue I am facing is how can i can bring in a fabric solution based on VCS to cater to the FCoE and iSCSI traffic requests originating from the hosts in Rack(s) A towards the storage gear hosted on Rack(s) B...
Much appreciate if you could please help me in designing a scalable solution in for this context...
01-05-2013 07:05 PM
just adding to the above in total there would be 40 Nos of VDX 6720's deployed in ToR formation across 2o Racks and 20 Nos of VDX 6730's deployed in 10 Racks in ToR formation and there would be max of 24 hosts per rack...
01-05-2013 07:56 PM
--->>> Much appreciate if you could please help me in designing a scalable solution in for this context
I'm sorry, but i don't have mapower at the moment to help with such Design.
maybe is another user here can help with.