06-07-2016 12:47 AM - edited 03-08-2019 06:06 AM
Hi,
I may have to design/build a very small (single rack) data centre presence in a location that is remote to our existing data centres.
I'm considering deploying vxlan in the primary (existing) DC and in the potential new one. This is for L2 adjacancy, and scalability in the long term.
All the designs I've seen so far use a spine leaf topology, or in some cases, the traditional hierarchical model. My question is, is it possible to deploy vxlan with only two nexus 9k switches?
The primary DC would have 9k leaves with a 5k spine, but I don't think I will be able to get the budget for a spine layer in the very small DC.
Is this possible? Would the multicast RP be a problem? Would it help if the RP were on separate devices (such as an ASA, which will be available)?
Thank you for your help.
06-07-2016 01:10 AM
Hi
5k cant be a spine , if its in aci mode it using vxlan it needs to be a 9336PQ standalone or a 9500 running a 9736PQ linecard, you will need a uc220 or 240 apic controller as well for the fabric
06-07-2016 01:20 AM
I don't need ACI just yet. I was under the impression that any router or l3 switch could be a spine, as long as ACI is not a requirement. Is this correct?
06-07-2016 01:48 AM
Only nexus 5600 supports vxlan if your not using ACI , not all platform's support vxlan yet
06-07-2016 03:46 PM
I realise that other 5k models don't support vxlan, which is fine. I was only thinking of using a 5500 as the spine layer, which is not vxlan aware.
I had a look at this blog:
https://comeroutewithme.com/2015/05/16/nexus-9000-vxlan-evpn-vpc-and-vxlan-routing/
It says "the spine switches are just doing basic routing — they see they have a packet that is destined for some leaf switch and route the packet along its way".
Unless I misunderstand, this means that they don't have to support vxlan specific features, just route the UDP packets that come from the VTEP on the leaves. Is that correct?
Would two 9k's support vxlan in a very small deployment without having this spine layer?
06-08-2016 07:41 AM
Yes thats correct the spine knows nothing of the vxlan the decapsulation is done between the leafs , i cant answer the last question as i only have a lab in this myself recently setup until our poc is complete and its running full fabric 2 leafs/spine and apic in full aci just for testing , i have not attempted to set it up with just 2 devices to see what happens
Your best off posting this in ACI section there are internal cisco ACI guys answering questions there they should be able to answer that if no one else does here in this section
06-08-2016 03:38 PM
Thanks, I appreciate your help.
03-20-2018 02:14 AM
Hello Luke,
Very Small VxLAN Deployment
A common deployment of a spine and leaf architecture I have seen is to use Cisco Nexus 5548 or 5596 switches with a Layer 3 daughter card as the choice of spine switch. Drawbacks : what is commonly missed in the specifications sheet is that once you add the Layer 3 daughter card into one of these switches, the Layer 3 forwarding drops down to 160Gbps or 240mpps This is a huge performance hit and is definitely not sufficient for a spine switch at a large scale.
11-25-2018 12:01 PM
Hi Luke,
Did you get the answer about your question? Is that possible to built a minimum VXLAN deployment with 2xNexus 5600 or 2xNexus 9300?
I read in some where it says that in VXLAN deployment, if Nexus 5600/9000 configured as LEAF, it will not be able to work as multicast RP. Cannot find the document anymore.
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